Posts for SpaceColonizer


Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
as ktwo said. i offered co-authorship before submitting. so i'm fine with it. i'm not an active enough member of the TASVideos community to know the proper etiquette. nothing in the TAS is copy/pasted from other movies. I do credit sources in my author's notes. but ktwo is largely responsible for the improvements to this TAS. These new updates wouldn't be happening if ktwo hadn't reached out. the only reason i'm the "main" author is because having me optimize the changes was easier than learning how to work with the sub-pixel mechanics. unless etiquette demands otherwise, i say add ktwo as co-author.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
https://tasvideos.org/UserFiles/Info/638431845236301656 here is updated TAS, 14 frames faster than the submitted movie. 12 frames from new bomb order in PR 2 by ktwo. 2 frames from new bomb order in PR 3 by me, inspired by ktwo's PR 2 routing.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
don't publish yet. ktwo came up with a new improvement for room 2. will be working on it this weekend.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
First sub 6 minute (TAS timing) MBJ 'Any Ending' TAS! Thanks in large part to ktwo for discovering a way to clip into the ledge on a totem pole, and coming up with whole new route for Palace Room 15. SUMMARY OF CHANGES FROM PREVIOSLY PUBLISHED TAS -Beginning: Got rid of two jumps in the last portion of the screen. This -20 points helps with lag reduction later in the run. -Palace Room 2: Improved routing developed by me. Saved 48 frames. Actually had this improvement a while back but started a 'Best Ending' TAS and was waiting to see if there were any more helpful discoveries from that project. -Shortcut: Collect four more bombs for +400 points for lag reduction benefits later. Totem Pole clipping discovered by ktwo saved 8 frames each for the three passes through the sphynx room. Total 24 frames saved. -Palace Room 11: Collect the p-button to turn mummies into coins. The only real benefit of this is that one of the coins is collected for +200 points for lag reduction benefits later. -Palace Room 15: Whole new route developed by ktwo saving 27 frames. Points manipulations earlier were necessary to make B-button appear at the moment of the first frame jump so that it could be collected later without causing extra lag. -Palace Room 16: Used to be 1 lag. Now there are none.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
Submission notes edited. They now go into more detail about the knowledge and tactics that made this level of optimization possible. Asked Samsara to delay publication so that the notes were there already at time of publication. Ready for publication. Thank you for your patience.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
https://tasvideos.org/UserFiles/Info/638252518783286140 alright. hope this is the final improvement. 70 frames faster than the previous file. There are a few different sources for these improvements: -Learned how to manipulate sub-pixels to get more efficient movements. Sub-Y pixels and "Sub-Gravity"(that's what threecreepio and I have been calling it) carry over from stage to stage so you can set yourself up for faster falls/jumps at the beginning of a stage by altering inputs at the end of the previous stage. You can also manipulate things mid-stage to a degree. -Figured out how to make the mummy in stage 10, that used to be in my way, fall through the platform. i still don't understand the actual mechanics of how this happens, but what seems to be happening is that conditions have to be right on multiple frames for the clipping to complete. if the mummy clips a little bit in on one frame, it can easily "snap" back up to the surface of the platform on the next. but with a little wiggling on the right frames you can continue the clip. -ktwo came up with a routing improvement for stage 11. very nice. haven't tested it in RTA yet because i wanted to finish the TAS before getting back to the grind. -Some minor lag reduction strategies. Hard to pinpoint exactly where i did what (should have kept notes on it), but usually just collecting a bomb on a different frame can avoid lag so not too much is happening at once. this can easily be done without slowing down the routing when collecting bombs while jumping or falling by shifting to the side of the bomb while passing it and shifting back into it on the desired frame. it's a small window of possibilities though, and sometimes isn't enough. I'll begin writing up detailed author's notes for the submission. and go more into all these tactics, but here's a summary of the frames saves for each stage compared to my previous file: Beginning Screen: No changes since the time to complete is based on horizontal movement. Stages 1-4: -1 frame each thanks to sub-pixel manipulations. Stage 5: -6 frames thanks to sub-pixel manipulations. Route was new last version so had more potential for optimization. Stage 6: -1 frame from sub-pixel manipulation. Stage 7: -2 frames from sub-pixel manipulation. Stage 8: -4 frames from sub-pixel manipulation. Stage 9: -2 frames from sub-pixel manipulation and lag reduction. Stage 10: -11 frames from sub-pixel manipulation and getting a mummy out of the way by forcing it to fall through platform. Stage 11: -24 frames from sub-pixel manipulation and routing improvement thanks to ktwo (18 of the 24 saved frames). Would have been -26 but 2 lag frames were added. Stage 12: -4 frames from sub-pixel manipulation and lag reduction. Stage 13: -3 frames from sub-pixel manipulation and lag reduction. Stage 14: +0 frames. Would have been -1 from sub-pixel manipulations, but a lag frame was added. Stage 15: -1 frame from sub-pixel manipulation. Would have been -3 but 2 lag frames were added. Stage 16: -4 frames from sub-pixel manipulation and lag reduction. End Screen 1: -3 frames from sub-pixel manipulation. End Screens 2 & 3: No change. End Screen 4: -1 frame from sub-pixel manipulation. NOTE: I have an alternative version which collects the B button on stage 8 in an effort to adjust the lag situation. Unfortunately it ended in a net zero change.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
currently working on more improvements.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
Samsara wrote:
Do you intend to continue working on this? Just making sure I don't judge the run before it's fully done.
to be perfectly honest, i don't know. i hadn't anticipated falling this deep down the rabbit hole when i started this project. let's hold off maybe another week. threecreepio and i were starting to narrow in on how a few memory locations that carry over from room to room can effect things. if we can figure it out we might be able to start each room with ideal conditions and squeeze out some more frames.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
https://tasvideos.org/UserFiles/Info/638241392787822463 New TAS update that's 52 frames faster than previous. Found new routing for stage 5 and a few improvements along the way. Also threecreepio discovered a pause strat for the end game area that saved a good chunk of frames. PAUSE JUMP EXPLANATION In the palace rooms that take up most of the run the B button is just another jump/hover like A and you can have them pressed on consecutive frames to do a quick jump and hover (if that would be useful). But on the other types of screens, B is used to activate mighty coins for power ups, so the fastest jump+hover is "A, blank, A". This kind of jump is needed for quickly breaking the blocks and chest during the end game sequence. But threecrepio discovered that if you "A, Start, blank, A+Start" you will come down for next jump quicker even though the inputs take 4 frames instead of 3. IMPROVEMENTS OVER PREVIOUS TAS Palace Room 2: Saved 3 frames that started with cleaner movement in the upper left (not the beginning, when coming back up). Palace Room 5: Saved 33 frames with new routing. Different first bomb with a last bomb that's closer to the exit. There is a more RTA viable version that uses this same first/last bomb. Palace Room 7: Saved 1 frame with a cleaner clip. Palace Room 14: Saved 1 frame, not exactly sure how. Palace Room 15: Saved 1 frame, not exactly sure how. End Screen 1: Saved 4 frames with pause strat. End Screen 2: Saved 4 frames with pause strat. End Screen 4: Saved 5 frames with pause strat and a faster exit.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
A best ending TAS is definitely something I will be happy to work on in the future, but not at this time. TASing is a part of my process when I'm trying to learn about games in the hope that it will benefit me in RTA runs. That's why I started out doing things on my own without looking at pre-existing TASes, I learn better that way. The first TAS I submitted to open this discussion was already 15 seconds faster than the canceled any% TASes of the past, and my next submission after realizing that holding down accelerated falls was 10 seconds faster than that, so I figured I was already doing things right and didn't need to look at anyone else's work. Obviously I was wrong. The routing, strategy, and tactics I learned from the other TASes shaved off another 4 seconds. Right now I'm working on trying to get a sub-7 RTA run in any% and the work I've done has certainly informed my performance. When I'm done with that I'll move on to a new project. Some day further down the line I'll return to the game to get a time in the best ending category. When I do that I will TAS a best ending run to help with my learning process.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
https://tasvideos.org/UserFiles/Info/638237798222671732 Alrighty. Here's another update to the TAS, which is 263 frames faster than the previous version. Big thanks to ktwo for pointing me towards some great resources. Most of the improvements come from adopting some routes and strategies from Hisatoki's best ending TAS and the "No Mighty Coin" strategy from Guybrush's TAS. Additional saves come from clipping into platforms by jumping or falling at the corner. Beginning Screen: Saved 27 frames by adopting G's No Mighty Coin strategy. Same time as G's because it's all unstopped rightward movement. G revealed the contents of many chests, but I only revealed the location of 5 Mighty Coins. 1: Saved 10 frames by clipping into, and jump boosting off of, the top platforms on the left and the right. Route was already 30 frames faster than H's; this time save makes it a total 40 frames faster. 2: Saved 2 frames by clipping into the middle left platform. Route was already 6 frames faster than H's; this time save makes it a total 8 frames faster. 3: Saved 21 frames mostly by adopting H's routing. Additional time saved by clipping into the lower left platform on the way up to collect a bomb early. H's route was 12 frames faster than my previous; this time save is 9 frames faster than H's. 4: Saved 23 frames mostly by adopting H's routing. Additional time saved by clipping into both of the left platforms. H's route was 16 frames faster than my previous; this time save is 7 frames faster than H's. 5: Saved 1 frame, not sure exactly how. Route was already faster than H's; this time save makes it a total 25 frames faster. 6: Saved 6 frames by clipping into the right platform. H's route is still 13 frames faster, but he isn't warping to the next Palace Room and therefore isn't restricted by first/last bomb considerations. 7: No frames saved. Clipped into left tunnel which should have saved time but didn't. Room is skipped in H's "best ending" TAS. 8: Saved 1 frame by clipping into the bottom right platform. Room is skipped in H's "best ending" TAS. 9: Saved 20 frames by dropping faster and collecting the missed bombs at the end. The faster start almost made it so that a mummy would be in my way and slow me down but I managed to force the mummy to fall through the floor. The only thing I manipulated to make this happen was Jack's x-coord, but there must be other factors. Room is skipped in H's "best ending" TAS. 10: Saved 6 frames mostly by adopting H's routing. Additional time saved by clipping into top left and bottom right platforms. H's is still 6 frames faster because at the end two mummies get in my way. In H's TAS the mummies stay on the top right platform because they have longer drop timers. This is either because H's TAS was done with Japanese ROM or because he skipped 3 Palace Rooms. Could save a lot more frames if I could safely walk under the mummy near the end. Getting the earlier mummy to pass through the floor so it wont be in my way would do the trick. 11: Saved 24 frames by clipping into and jump boosting off of the center top platform. The faster left side got me to the right side fast enough that the last mummy spawned on the left and wouldn't be in my way. Also clipped into upper right platform. H's route is still 10 frames faster, but he isn't warping to the next palace room and therefore isn't restricted by first/last bomb considerations. 12: Saved 25 frames by adopting H's routing. H's route was 14 frames faster than my previous; this time save is 11 frames faster than H's because it has less lag, although I'm not sure why. 13: Saved 12 frames by clipping into the middle left platform so I could collect the bombs while going up without having to land on the platform. Route was already faster than H's; this time save makes it a total 67 frames faster. 14: Saved 30 frames by adopting H's routing. H's route was 26 frames faster than my previous; this time save is 4 frames faster than H's. 15: Saved 33 frames by adopting H's routing, except for the jump boost at the end which kept being 1 frame slower for me. H's route was 29 frames faster than my previous; this time save is 4 frames faster than H's. 16: Saved 1 frame by clipping when exiting of the central box. Route was already faster than H's; this time save makes it a total 25 frames faster. End Screen 1: Saved 2 frames, not sure exactly how. End Screen 2: Lost 2 frames because of the No Might Coin strategy. End Screen 3: Saved 21 frames by jump boosting up to the door. End Screen 4: No change.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
OK, after looking at the older TASes I see lots of way to improve mine. i will continue working on it now. i see what you mean about jumping across the platforms now. and the beginning/end changes will be great.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
hmm. looks like I have some old TASes to look at and will make improvements to my submission. for the record, i've never seen them before, so any similarities to those TASes is coincidental, or because msm_smtn copied those starts. my TAS is mostly based on my RTA routes, which were heavily influenced by msm_smtn's WR. RE: dropping over the corner of platforms instead of walking off. my TAS already accounts for that, if there is anywhere that i walk off the edge still it's probably because getting to the floor ASAP wasn't the priority at the moment and left/right was more important. walking off the edge is basically a free hover. i just tested going over the left side of the bottom right platform at the end of 3... going over the right is 4 frames faster for me. i don't know if the TAS you want me to look at does it in a special way but the way i'm programming them: right side is faster. going left requires that you go horizontally to the left, then straight down to the floor (aka orthogonal movement only). by going right a little instead, you only have to go straight down a little bit, then turn left while falling. that diagonal fall is more economic. i'll look at hisatoki's routing for 12, but i have a possible improvement already that involves going across the top first for the bombs on the right, then going back left to continue the normal routing. this avoids going down/up an extra time on the right side at the end. but it only saved a frame when i tested it. maybe hisatoki does something completely different. interesting to see how hisatoki did 14 and 15, but i already think i understand what you're saying about pillar jumping in 15. as far as the beginning/end. i honestly thought two mighty coins were required to open the chests in the tower of chests without going all the way to the top, i'll see what hisatoki does. that'll be a big time save. just for the sake of pointing out how similar my TAS routes already are to my RTA routing, here are the major differences from the TAS: 1: i fall straight down at the start instead of the weird around the platform drop. 2: i don't do the frame perfect first jump in RTA. 3: i don't just collect the bottom bomb on the left, i go all the way up in RTA. doing the u-turn to get the top four bombs in the middle is very efficient in TAS but too slow in RTA. 4: when i make another pass at the TAS i'll improve some things in 4 that make it closer to my RTA routing. 9: TAS does a few weird things i don't risk doing in RTA, but i definitely use that beginning to require less jumps at the end. 15: this WAS what i was doing in RTA, but threecreepio just helped me develop new RTA routing that's a little faster and a lot safer. unfortunately the new RTA strategy is not faster in TAS.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
i'm not sure if "E" and "C" are doing what the notes say they should. at the very least they aren't updating every frame. C seems to only change when a bomb is collected. not sure if i'm misinterpreting the numbers. but here's when the mummies spawn in the closer spawn location instead of the farther in the second version of my TAS. 3rd mummy in 4. 1st mummy in 6. 3rd mummy in 8. 4th mummy in 10. 2nd mummy in 12. 3rd and 4th mummies in 14. interesting that it only happened in even numbers rooms. possibly just a coincidence. only 2 mummies had a chance to spawn in room 2 with the way it's currently routed, but i edited it and the 3rd mummy would have been close not far. a mummy falls through floor on 14. (there is a mummy directly above it that is also about to fall. may be related) the drop timers are interesting. they don't count down at a consistent pace. the number of frames it takes to count down to the next value changes. looked more into jellyfish/metroid movement. looks like their initial movement is based on where jack was when it was "selected", not where jack was when it starts moving. after that it moves towards jack when ever it hits a barrier, speed determined by how far away jack was at the moment of the bounce, moving faster if jack is farther away. jack touching walls has no effect that i can see. looks like if they hit a barrier but want to continue moving in the same direction (because jack is still in that direction), they will reflect their bounce.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
(before looking at your notes notes) -yeah so I've figured out a few things: where they spawn is effected by where jack is (at least horizontally, not sure vertically). it seems like they usually spawn on the opposite side of the screen from jack, but not always the case. (this is how stingrays spawn in jaws, opposite side from player) -they usually seem to walk in jack's direction after spawning, but i've seem them behave differently in rta. -the main thing i don't understand is when they'll fall off an edge instead of turning around. it's not number of "bounces/turns", it's something else. -what type of spawn they become after landing on the floor is determined by how many times you jump/hover, which is tracked globally. the jumps are kept track of in memory value FA (thanks to threecreepio for finding that for me). (after reading your notes) -oh wow, the location is based on what's farther away not what's on the other side? interesting. will have to play around with that. -hmmm. drop timer. interesting. will have to look into that. sometimes mummies drop through a floor instead of landing on it. not sure if that's a clipping glitch or if it's based on where jack is somehow. -so i has some notes on the enemy movement patterns. we see the same behavior on everyone except on the "jellyfish" (which i call metroids). what i think they're doing is they move towards jack whenever THEY bounce off a wall, how fast they move is determined how far away jack is. so if jack is close, they move slowly towards him on a bounce, and if he's far they move fast towards him when they bounce. i've never witnessed they're behavior changing because jack was the one that bounced, but i could look out for that now that you mentioned it. -hmmm. threecreepio and i were already saying mod6 just for short hand, but i hadn't considered the rollover situation because of 255.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
https://tasvideos.org/UserFiles/Info/638231723649985300 here's the user file link to my updated TAS. about 10 seconds faster than previous. also revealed the location of two more medallions. only two are needed for the end game in the TAS and in the WR. but i use a safety strategy that involves getting 4 coins, so thought it would be neat to show their locations in case anyone watches the TAS for strategies. unless someone can point out any major oversights i will stop working on it.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
ktwo wrote:
There are two important game mechanics that you have overlooked: - By pressing down on the d-pad, you descend faster than freely falling. Can be used in pretty much every room. - The horizontal movement has acceleration. You should therefore avoid pressing against walls when falling and instead build up speed during the fall so you can clear the obstacle at max speed. Can be used in e.g. the right side of room 3 (and probably other rooms as well). I'm very interested in seeing a TAS of this category. However, there is in my opinion not enough complexity in this game to allow for basically even a frame of known improvement. If you decide to redo this and squeeze out the last frames, I'll try to take another look.
hey. i've been working on some improvements already. i figured out the pressing down thing on my own. major part of the improvements i'm getting. replacement TAS should be about 8 seconds faster. as far as the horizontal acceleration and pressing up against barriers, i test multiple possibility whenever i'm in those situations. for some reason pressing into a barrier can sometimes get you to your destination sooner. some sort of clipping maybe i think.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
sorry for any hassle this causes the publisher, but I have another improvement to the TAS. In the future I'll be sure to spend some more time on a TAS before getting the submission process started. This was the first time i wrote a TAS knowing I would submit it so i got a little ahead of myself. https://tasvideos.org/UserFiles/Info/638069963297013797 -greatly improved varan fight, finally figured out how to get critical hits when i want them. -saved a few frames on the death warp (including the set-up), but more importantly i grab three life capsules instead of two. this is important because i realized that the rate that your blat attack recharges is linked to current HP, so the extra life will make some fights later faster. -time saves on hedorah, gigan, and ghidorah thanks to faster blast recharges. -added a death warp to the final stage. i'll begin editing the author's notes to reflect the changes in the TAS.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
ThunderAxe31 wrote:
Oh so this password affects the enemies for the first planet only?
yes
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
ThunderAxe31 wrote:
SpaceColonizer wrote:
-Category rules require you to defeat all enemy Kaiju before exiting the planet.
Just to be sure: is this required in-game to advance to the next planet, or is this a self-imposed goal?
SpaceColonizer wrote:
Since the special content from the DAM password is only on Earth, exiting Earth is the suitable endpoint for the category.
So, would killing them again be exactly the same as on Earth?
self-imposed (although you would still have to defeat most of them to clear a path to the exit). i don't understand the second question. killing them again? the enemy kaiju return to their normal power levels on the later planets and they aren't all present together again until planet-x,.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
In Any%: -Godzilla and Mothra start on Earth at lvl 1 each. The level cap on Earth is 3, and slowly increases on each planet until lvl 16 on Planet-X. -The only Kaiju on Earth are Gezora and Moguera at their lowest level. Additional Kaiju are slowly added on other planets and established Kaiju increase in power. -You are not required to kill enemy kaiju. You only need to get one player monster to the exit tile of each planet. -End Of Run: End Credits after occupying the exit tile on Planet-X In Destr0y All M0sters(DAM): -Godzilla and Mothra start on Earth at lvl 8, already beyond the lvl cap of 3. -All Kaiju are on Earth at their maximum power (as they would be on Planet-X). -Category rules require you to defeat all enemy Kaiju before exiting the planet. -End of Run: Exit Earth. Since the special content from the DAM password is only on Earth, exiting Earth is the suitable endpoint for the category. I don't think reaching a more definitive end screen after Earth would increase the branch's viability. For context, there are short passwords that take you to any planet (MAR, JUP, SAT, URA, PLU, NEP, X STAR) but with Godzilla and Mothra starting at lvl 1 instead of what you'd normally get them to during the coarse of a full run. We use these passwords for an individual level leaderboard. If I made a TAS of a run starting with the X STAR password, it would reach the end credits but I don't think that would be a more valid branch. Also those passwords don't change the content of those planets beyond starting Godzilla and Mothra at lvl 1.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
feos wrote:
This mode still has a proper ending if you play further, right?
yes. i've done an RTA run I called "DAM All Planets" before where I continued a DAM run through the other planets and kept the requirement to defeat all the enemy kaiju (a requirement that doesn't exist in any non-DAM categories). But nothing about the rest of the game changes except that godzilla and mothra got a head start on levels (there's a level cap on each planet so they don't start earning experience again until saturn or something). We've never added it as a category on SRC. The same is true of the gh1d0rah password game: the only planet that's different is Earth so that's all we play in those categories.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
ok. finished the author's notes. here's a 285 frame improvement over my file I originally submitted. biggest area of imporvement was baragon fight. realized i could get more punches in if i forced him to jump for better dps. death warp is also improved by about a second. smaller time saves in other places as well. https://tasvideos.org/UserFiles/Info/638056538424500106
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
Working on some improvements to the TAS that I thought of while writing the Author's Notes. Going to work on it a few more days.
Experienced Forum User, Published Author, Player (205)
Joined: 9/11/2021
Posts: 37
OK. Implemented your suggested routing. Cuts off 196 frames from inputs. redoing the rng manips wasn't so bad. But when looking into fastest last input possible, found something interesting. First off... if you leave too many windows undestroyed, there are too many solider guys throwing grenades, and since you're not moving they will mostly hit you. These will break you down before the dynamite guys come out. So destroyed three buildings, with the last being the big one and that seriously reduced the incoming damage. Also since the route stops me on the top of a building, I was out of reach of the cop car until a missile knocked me down, so avoided that damage for a long while. then here's where things get really interesting... somehow the game glitched so even when on the ground i was immune to the cop car! without that, lizzie would have most likely been damaged to death before the dynamite guys finished the job. it does take a LONG time for those dynamite guys to start coming though. This cuts off another 214 frames for last input. I'm conflicted. I prefer that the buildings be destroyed be player action. but that surprise "f*** the police!" glitch is interesting. I think my preference would be to submit the updated version with your improvements, then just put in the notes about the fastest last input. It's as easy as "truncate from this frame to see X Y Z". EDIT: here's new TAS: https://tasvideos.org/UserFiles/Info/638050330128198356 truncate at 315159 to see the glitch