Posts for Maru


Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
I am appreciative of all of the positive feedback concerning the TAS! Happy New Year to you guys too! fcxiaopengyou: 1. Even if the number of instructions is sufficient, a useful object won't appear in Cutman. I will mention that when making this TAS, we discovered that an address can be changed in Bombman that will prevent Wily 2 DelayObjectGlitch from working altogether, but I doubt this is the issue here. 2. Score must always be manipulated to shorten the score countdown animation at the end of each level. 3. I looked at the video and cannot seem to reproduce this glitch. It seems like the health counter is only filling up to 4 instead of 28 like it should, but even when I tried to set everything up, it still filled up to 28. If someone is able to reproduce it, it will certainly be worth looking into, but I am a bit skeptical about the legitimacy of this video as well, considering it is unlisted. I also do not know which emulator was used in the video. It takes 4.8 seconds to zip through the wall just before the boss fight. Running with the water current is approximately 5.5x slower than zipping speed (which is about a 22 second difference alone). You could theoretically use some of those magnet beams to perform zips through that corridor (you'd have a few to use), but I doubt it would be fast enough. It also takes longer to get to that corridor instead of just simply getting into the wall and zipping freely. The boss fight in our TAS is not as long as people think (about 18.6 seconds including the 252 frames of lag needed to begin the boss fight). If this glitch can be reproduced, the boss will be no more than 14.4 seconds faster. While it is interesting, I highly doubt we can save time with this. Thanks for sharing.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
If anyone is interested in improving this movie, please see my test file in my judgement note. The general rule is that it is best to stay in the air for as much as possible.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
https://pastebin.com/NkA9NZWt SMB3 warpless has a couple of routing improvements that I have decided to share here. I have also successfully created a TAS route for the "All Forts" category that GDQ events have popularized.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
I have a feeling that I need to add some stuff here due to questions that people have asked me. Rejecting a run like this would essentially be placing a ban on DPCM workaround abuse. DPCM workaround abuse can be used to unleash a payload in 1-1 of SMB3, which does include being able to load the credits (https://www.youtube.com/watch?v=7CgXvIuZR40). If a TAS was submitted that reached the credits in 1-1 taking advantage of the same hardware glitch that this TAS uses, there would not be nearly as many complaints. Likewise, people wouldn't be raising many complaints about abusing the DPCM workaround to reach the credits in 1-1 of SMB2USA. https://www.twitch.tv/dwangoac/clip/TransparentMushyAnteaterMVGame Now, regarding having to differentiate game end glitches from each other, the only two ways to truly differentiate them are by how they look and other technical attributes. Obviously, for SMB3, these two game end glitches look very different, but what about for other games? SMW's current game end glitch run uses a jump to Open Bus, which is a region in the SNES that allows for several different possibilities. It's not that much different from the RTA record for this category (https://www.speedrun.com/smw/run/zp0q5rvm), but the latter does not use a controller payload to assist with reaching the credits. We have to go back in time to find a TAS that does not utilize the Open Bus region of the SNES (http://tasvideos.org/2380M.html) and even more back in time to find a TAS that does not take advantage of a jump to controller data (http://tasvideos.org/1945M.html). Both of these game end glitches look different enough from the current one, I'd say. The faster game end glitch obsoleted the slower one. If we lumped TASes taking advantage of this DPCM workaround into their own category, then what do we do about game end glitches taking advantage of the Open Bus region and TASes that rely on jumps to controller data? These two things are quite powerful too. Do those go in their own separate category? Remember that having meaningful publications is one of the goals of our site. The recent ALTTP game end glitch takes advantage of a jump to controller data, while the old publication did not. Should those have been separate all along? I think that's undoubtedly one big mess. We shouldn't be placing arbitrary limitations on the DPCM workaround because then, we could look at placing arbitrary limitations on the use of other things that have already been used to create faster game end glitch TASes. The DPCM workaround is just another thing that lets us push games to their limits. Admittedly, I would be concerned if every NES game was vulnerable to the glitch to the extent that SMB3 is, but I did make it clear in the judgment note that only a handful of games would be vulnerable to a quick credits warp like this. Overall, when everything is boiled down, there's no clear way to differentiate game end glitches that promotes organization and efficiency, but when the goals of the TASes are looked at (beating the game as fast as possible), this game end glitch TAS is going to obsolete the slower one.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
Alyosha wrote:
6 sided dice roll generator I found a 3 frame time save in the warpless run in 4-2. If you run a bit before pressing down for the super jump, the sliding momentum moves you quite a bit further while building up the jump power. The picture above is the frame I press down. Unfortunately fry guy desyncs due to ??? so it still needs work.
Good find. I have some small improvements to 1-1 and 4-1 that work for both the SMB2USA warps and warpless categories. SMB2USA in general is hard to optimize due to the oddities of subpixel carryover and peculiar character physics. The princess only TAS has more improvements though. There's a faster route through 1-1 with the princess and some other optimization improvements. That might be a branch worth TASing if anyone is up to it.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
I'd like to raise a few questions here because this TAS has mixed feedback so far. I was very brave to claim this TAS for judgment. This type of TAS is something that has rarely, if not never, been seen here before. With regards to console verification, at the moment of this post, this TAS has yet to be console verified. However, it may be possible with some tweaking of the .r08/.r16m file. What do we consider as gameplay? This TAS is going to force us as a community to reevaluate this concept all together. In this submission's case, all of the work is done at the title screen. Do we consider what is being done there as gameplay? Note that TASVideos follows a strict timing rule for most TASes. It is from power-on until the last input. SMB3 follows the standard timing guideline. However, this is not true for the SMB3 RTA commnity. For the record, the RTA community for SMB3 has decided that the time starts when the start button is pressed to load the World map. Now the question is, based on those timing guidelines, which guideline better constitutes gameplay? This is not the first time when a game end glitch TAS that does not make use of a controller input payload to finish the game was obsoleted by one that does. The most recent example is Tompa's ALTTP game end glitch run, which was obsoleted by fmp, total, and Yuzuhara_3 using a four frame controller payload. Another example that comes to mind was Masterjun's SMW2:YI game end glitch TAS, which used a payload in 2-2 with multiple controllers to obsolete the old warp glitch TAS that used a glitch to warp straight to the last level of the game. While the same warp glitch was used in both TASes (Masterjun used the warp glitch to go to 2-2 instead of the autoscrolling 6-8), one of them used a controller input payload to reach the credits while the other did not. What separates those two TASes from this one is that some setup past the title screen was required to get the game end glitch to work. There were gameplay improvements in both TASes, coming from either a new route or better optimization. That brings me to my next question, which is: Does this TAS have any gameplay improvements compared to the published run? To answer that question, you will need to consider the first question too. This next question is unrelated to the others. If this submission is accepted, should it obsolete the published run based on the fact that the published run was thoroughly outperformed in terms of time? At TASVideos and especially with games like SMB3, new categories of games have to show significant differences from the existing categories in order to be accepted. This brings up the next question. Is this type of game end glitch significantly different from the published game end glitch TAS to warrant the creation of a new category? This is relatively unheard of for game end glitch TASes, but we all have to be open to possibilities. Another question does come to mind looking at this TAS from a viewer's perspective. Since the ending is almost instant, do viewers see some similarity or resemblance to entering a password in-game to skip to the credits? No passwords are used in this TAS, but whether people find it similar to using one is a question that needs to be answered. I could see arguments that any sort of game end glitch could look like using a password to skip major portions of a game. However, it is worth mentioning that prior to the credits, Mario's face is not seen in any way, shape, or form. All a viewer sees is the curtains prior to reaching the credits. I would love to hear people's thoughts on these questions. Remember that there is no right or wrong answer to what I have asked here. All of these questions are open-ended. The decision for this TAS is going to be a very difficult one to make. The goal here is to set a good precedent for future game end glitch TASes, especially ones of this type. I do not want to set a bad precedent for future game end glitch TASes here.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
I was made aware of this submission, and I thought I would come here to provide a little bit of context. For reference, left+right allows for faster acceleration and some deceleration tricks that would be otherwise impossible. Therefore, it is banned in the real time speedrunning rules of the game. Many years ago, Andrewg asked HappyLee to make an RTA Rules TAS, or one that does not press left and right at the same time. HappyLee made one beating the game in 4:57.56. It matched the published warps TAS in all regards, but it lost 15 frames in 8-4 due to the no left+right restriction. It was fairly well optimized, but three months ago, I decided to take a look at it for myself and saved a frame in 8-4, bringing the time down to 4:57.54. It ended up being a major milestone for SMB1 TASing, as an editor from Kotaku interviewed me about it. However, it is worth remembering that the purpose of this TAS is mostly to have a "fair" comparison to real time speedruns. There are not many visually noticeable differences between the published warps TAS and the RTA Rules TAS, so it's not worth putting a TAS like this on TASVideos. It seems like zdoroviy_antony decided to make an SMB1 RTA Rules TAS of his own. There are several people in the SMB1 community who want to make their own version of the fastest TASes. For some unknown reason, he decided to submit this TAS with my name under the authors column, which I told Mothrayas to remove. TASVideos is not like SRC/speedrun.com where there is a leaderboard for each game. All TASes have to beat existing records, and new categories must show significant differences from existing ones. I hope I have made my point clear.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
World 9 and Worlds A-D can be featured in the FDS TAS if it starts from SRAM, but that is usually not allowed. What I will say is that this movie is unoptimized. It looks like it was recorded on top of HappyLee's SMB2J warps TAS because the movie file has HappyLee marked as the author ("author HappyLee" in the metadata). While the movie file says 72632 rerecords, 70167 of those rerecords come from HappyLee's warps TAS. In addition, it is clear that this TAS was not done with memory watch. I noticed some clear improvements besides the Bowser fights. The 1-3 mushroom grab, 1-4 fire flower grab, and maneuvering around the blocks leading to the Bowser fight can all be improved, and this is only considering World 1. The fast accelerations in castle levels are also suboptimal. I do think this game has potential, but someone needs to make a well-optimized TAS of it first. Here is a TAS of Worlds A-D done by chatterbox which looks pretty good: https://www.youtube.com/watch?v=2t8CK3SivAc I would recommend sharing WIPs so other viewers can offer feedback. It can greatly increase the quality of the final submission.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
GJTASer2018 wrote:
Mitjitsu wrote:
Surely now we won't be seeing any improvements to this in the future.
I would not be too sure about that - that's what was said about the original Super Mario Brothers before this happened. :)
SMB3 doesn't have the level completion framerule that SMB1 does, but it has several other RNG-based framerules that prevent small improvements from helping. HappyLee's one frame improvement to klmz's TAS came from 8-4 because that's the only level where every frame counts in SMB1. All other levels are subject to the 21 frame framerule. While there are minor speed improvements such as a one frame improvement to 1-2 and a faster route to push ahead of the screen in 4-2, none of those matter because of the level completion framerule. The framerule also made finding new improvements very difficult in the SMB1 warpless TAS. SMB3, on the other hand, does have an RNG-related Bowser pattern framerule in the last level of the game. We need to start the Bowser boss fight on a frame that causes Bowser to hop six times before he makes his big jump. Bowser can hop three to six times, and the six hop pattern is the least common Bowser pattern. The reason we need this specific pattern is that we want to defeat Bowser on the frame he starts his sixth hop. This allows Bowser to drop much lower upon being defeated, which saves approximately 50 frames. With a different Bowser pattern, the Bowser Quick Drop glitch won't work until after Bowser makes a big jump, which wastes time. In World 1, we have to time our level endings for the shortest overworld hammer brother movements. The hammer brothers can move once or twice, and an additional movement costs about 30 frames. In 8-Tank1 of World 8, we have to fight a boomerang bro at the end of the level. However, we need to start the fight on a specific frame in order to complete it as fast as possible. The boomerang bro can either move forward or back, and the chest won't spawn until all of the boomerang bro's sprites leave the screen. When the boomerang bro moves back, the boomerang that he throws leaves the screen faster, which saves time. The next optimal Boomerang Bro pattern would involve starting the fight five (5) frames faster than our TAS. While we know that World 1 can be completed four frames faster than our TAS, bad overworld hammer brother randomness after the World 1 fortress means that we would need to save an additional nine frames (overall 13 frames faster in World 1 compared to this TAS) in order to net savings of at least five frames after 8-Tank1. We would hope that being at least five frames faster after 8-Tank1 would give us better luck at the Hands. We had to delay multiple frames to manipulate the first and third Hands, something that the 2010 TAS by Lord Tom, Mitjitsu, and Tompa did not have to do. While the World 8 Fortress has a lot of moving parts to the level, acceleration framerules make this level very difficult to improve. There's an 8-frame rule for building up the P-Meter and an acceleration framerule that prevents Mario from accelerating every 8th frame. The 8-frame counter is controlled by $55d in this game. While we could enter the door to the second room of 8-F one frame faster, that did not save any time because of bad acceleration framerule luck. So, we delayed one frame to enter the door with a higher subpixel value ($74d = 240 instead of 176) because subpixels are carried over after levels. I had found a strategy that could've potentially entered the door two frames faster in the first room, but after quite a bit of testing, I was never able to get the subpixels to line up in order to not lose the time I gained on one of the jumps. The first room of 8-Fortress requires precise subpixel optimization in order to maneuver around the walls and roto-discs as fast as possible. Even after using cheats to enter the door two frames faster, the first star grab became problematic because of worse acceleration framerule luck. After grabbing the second star, the acceleration framerules still need to line up to regain P-Speed as fast as possible after hitting the P-Switch. On top of all of this, the next six-hop Bowser pattern would involve having to start the Bowser fight 25 frames faster than this TAS. With what we know about the mechanics of the game, that does not seem likely. On the other hand, we are very happy with what we achieved, and I think SMB3 still has a lot more TASing history left.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
A TAS of this game is in progress. Currently, we are in World 3, and at the moment, the TAS is over two minutes ahead of the current RTA record of 1:02:34 by JabemX. I may decide to post updates periodically on this project.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
Feos, thanks for the comparison. In World 1, I use a different powerup route than Nitsuja does, so I lose time in 1-3 because he grabs both the mushroom and leaf in 1-1. However, this route is faster overall. I had to delay a frame on the fortress to manipulate the hammer brothers on the World 1 map. Nitsuja's TAS did not seem to pay attention to hammer bro manipulation. While I was able to complete the fortress faster, it did not provide good hammer brother RNG on the world map. 8-T, however, did have an improvement. I had tried long enough resyncing the level after delaying level entry by a frame to see if I could manipulate the two-frame framerule into giving me the optimal Boomerang Bro pattern earlier, but I forgot that the Boomerang Bro disappears off screen faster when he is jumped on rather than tail swiped! :@ I was able to save 36 frames in 8-T from that, but some of that time was lost manipulating the Bowser pattern. Please see http://tasvideos.org/userfiles/info/52145446707527156 for the improvement. Please also fix the submission to reflect the improvement.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
There are some differences that would be present in the GBA version. I've never done any work with SMA4, but as far as I know, some of the skips that are in this and the NES TAS are not possible on the GBA version. This seems to include the 8-BC route and possibly other routes as well. I have also observed that the camera can be abused in SMA4, which is not the case with this version. It can be used to kill piranha plants like the one in 8-1 or the series in 8-2 without as much slowing down. Apparently, it can also be used as a quick boost when Mario bumps into a block. Some World 8 autoscrollers are shorter in the GBA version compared to the NES or SNES version. That means it is impossible to achieve 99 lives from it. While some people think autoscrollers are boring, I don't think the difference is noticeable as a whole because the GBA version has longer cutscenes. Other than that, I would need to investigate whether more minor glitches like the airwalk glitch, bob-omb glitch, and drop glitch in 8-BC are even possible on the GBA version. While I understand that some people think the game is too similar to the NES version and the GBA version, there are some strategy differences. For one, the 1-1 mushroom grab would probably look different (and slower) on the GBA version due to level differences. The layout of 1-F in the GBA version also allows Mario to get P-Speed earlier for flight, which is not ideal on this version. I think some of this boils down to preferences as well. Some people prefer the GBA version for various reasons, while others prefer the SNES version. I personally think the GBA version has worse music and annoying sound effects, but others may disagree. I don't know if the GBA version has this so-called "two frame framerule," but that is one of the main differences between SMAS:SMB3 and the NES version, and it makes manipulating luck (in this case, World Map Hammer Bro. movements) a lot more challenging. You can either manipulate luck by finding improvements or delaying frames. Finding improvements is harder because the RNG depends on what two frame window you are on, and delaying frames with this in mind means that you would be losing more time to manipulate luck. Overall, the SMAS:SMB3 version is faster than the GBA version. After we finished the NES SMB3 TAS, I wanted a change of pace, so I made this run thinking that it would be at least a few seconds slower than the NES version. Eventually, after some tight optimizations, I was able to surpass the time of the NES TAS. I did not expect to see that happen, and I definitely do not think others expected it either.
Active player, Experienced Forum User, Published Author (479)
Joined: 7/22/2018
Posts: 14
I am glad that people enjoy the TAS. I was worried that some people would think that this was too similar to the NES version, but it seems that people enjoy the use of left+right and some of the unique features of this movie, including the new 8-F route. I understand that some people want to see more of this game, but I think I need to weigh in on that. There is a very low chance that I will create a warpless or 100% TAS of this game. If someone wants to commit to the project, I am happy to watch WIPs or contribute some strategies, but it would be a difficult project. As much as left+right simplifies this game from a TASing perspective, the two frame framerule will make manipulating RNG (Hammer Bro. movements) on this game more challenging compared to the NES version. That would certainly complicate making an optimized TAS, especially considering the precise manipulation of Hammer Bro. movements that go into World 2 and World 3 of the warpless TAS. I don't think a 100% TAS would be as challenging to make as a warpless TAS, but the RNG manipulation issue still stands. In addition, there are some interesting strategies that are unique to the All Stars version. For example, there is a ceiling clip that you can use on 7-5 with a P-Wing that is impossible on the NES version.