Submission #7027: Reseren's GBA Metroid Fusion in 1:12:19.31

Game Boy Advance
baseline
BizHawk 2.3.0
259176
59.7275005696058
11734
Unknown
Metroid Fusion (USA, Australia).gba
Submitted by Reseren on 2/9/2021 2:38:49 AM
Submission Comments
Emulator is BizHawk 2.3.0 with mGBA core.
  • Aims for fastest possible time
  • Collects any items
  • Takes damage to save time
  • Abuses programming errors to save time
  • Manipulates luck
Unlike previous TAS's which prioritize in-game time, this TAS aims for the fastest real time possible. During some door transitions, the door will align itself by moving either up or down, wasting a small amount of time. By jumping through these doors, they can be aligned before entering the transition. For this reason, there are many places where jumping through doors is actually faster in real time than simply running through them. The total real time saved by doing this adds up to around 13 seconds. There is also a small route deviation when leaving Sector 5 which is faster RTA than the usual route.
Another improvement over past submissions is the use of out-of-bounds clips to force Samus into walls. If Samus is directly between a ceiling and a floor (or frozen enemy), she can clip out-of bounds into an adjacent wall by unmorphing and immediately spin-jumping. This allows Samus to ignore walls and traverse rooms very quickly that would otherwise be relatively slow. For example, the SA-X interaction in Sector 6 can be effectively skipped, saving over 15 seconds.
Samus can also clip out-of-bounds when jumping into a door transition. As she exits the door, she can turn around and morph just as the door is closing in order to clip into the wall. This is used to chain together spring ball jumps to ascend walls very quickly. There are only 3 places where this is implemented because it is usually not faster due to obstacles such as enemies in the way.
Picking up 3 missile packs is optimal; it allows for a faster nettori and nightmare fight purely from how much faster missiles do damage than charge beam. It also allows us to completely skip Yakuza's jumping phase by bringing it down to 0 hp in the first phase. Picking up more than 3 missile packs ends up being slower because of the extra time it takes to collect them. In total, 3% item collection is about 24 seconds faster than 0%.
Luck manipulation is a pretty big thing in this TAS. Getting ideal eyedoors and enemy patterns requires delaying room entry 2 frames at a time (because doors are only active every other frame). Sometimes if the room is very unlucky and does not cooperate, I have to settle for a mediocre pattern because delaying room entry ends up wasting more time than a good pattern would save. This was particularly annoying to deal with in the room right before the Yakuza fight; not only did I have to get good rng on the eyedoor, but I also had to get good luck with the space pirate patterns. In the end I had to accept a somewhat slow eyedoor because entering the room later would just make me lose more time in the end.
The most challenging (or at least time-consuming) part of this TAS was probably the Core-x fights after each boss is defeated. There is a specific spot in the room where the Core-x drop will float to after you destroy it. If you destroy the core in this "sweet spot", you can pickup the upgrade immediately. If it is even a few pixels off, the drop will float for a few frames before you can grab it. So in order to get a perfect fight, the Core-X needs to be hit with 2 missiles on the same frame while in this exact spot, AND have Samus positioned as far to the exit of the room as possible. Core-x fights alone took up about 15% of the total time to make the TAS, while they are only about 1% of the total run time.
I'd like to shoutout Biospark and Dragonfangs for baselining many of the strategies used in this TAS.

Samsara: File replaced with a version that trims several minutes of blank input. Also, judging.
Samsara: I initially had some concerns with the optimization in this movie, as there were a few sections that appeared to be noticeably slower than the two currently published runs of Fusion, however those concerns were alleviated by the author's response and some outside discussion with others more familiar with the intricacies of the game. This run is essentially directly comparable to the published 0% run room by room, except for the three collected Missile packs, and given the time save over it even with the routing and emulation differences (BIOS screen, the seconds wasted to collecting the Missile packs, extra lag on door transitions), it's definitely clear this is far more optimized than that run. Comparing to 100% would take an incredible deep dive due to the routing not being 100% comparable at any point, and due to the game's frame-based RNG, any time losses (including one I found very early on) are almost certainly unavoidable without wasting more frames to try and get a better pattern. In short, a proper comparison would likely take another complete TAS, and I don't feel like that's necessary in this case. The overall quality is high enough for acceptance, and the most noticeable "time loss" is easily explained by a trick involving jumping through doors to reduce transition time at the expense of IGT (the end sequence is a half second slower than the published 0% run in IGT, but it's faster in real time).
As for the category, the appeal of 0% is that it skips a Missile pack that's normally meant to be unskippable. The Japanese version even has a special low% end screen that triggers at 1% collection but not 0%, implying that this Missile pack is "mandatory". However, there's a bit of a problem with categorizing 0% in TASvideos terms, and this is the submission that proves it: It is way too similar to any%. Given that Fusion is infamously linear, it's extremely unlikely that a new route will be found to differentiate the two, so I see no reason to have them both published alongside each other. In fact, the "mandatory" missile pack that was such a novel skip for 0% is still skipped in this run (relevant section starts at 26:55 in the encode above, storing a shinespark to bypass that vine), meaning that the primary appeal of the 0% run remains here in this run. There's one last problem that contradicts what I just said, and that's the existence of a skip using memory corruption, cutting out the last half hour of the game. This means two things: One is that there's a third category for this game, since we consider the general scope of memory/save corruption/ACE to be a separate category that only competes with itself. The other is that, as far as I can tell, this would also technically serve as the 0% category. I'm not an expert on Fusion speedrunning, so I could be wrong about this, but it seems like the primary purpose of collecting the extra Missile packs is to kill Yakuza in one phase. This fight is in the last 30 minutes of the game that ends up being skipped by the corruption route, meaning that the extra Missile packs would likely take more time to collect than they would save over the rest of the actual gameplay. Granted, if a save corruption TAS came along instead of this one, it would still be a separate category alongside 0% (this is why I said technically serve after all), but the existence of the corruption route in general means that it's even harder to justify having 0% as a separate category at all, given that any% has always been faster and the only reason we haven't had an any% submission until now is literally because prior TASers just chose not to do it.
All that being said, I'm accepting this to obsolete the 0% TAS. It's more optimized, it's faster, and it still keeps the same trick that warranted 0% being a category in the first place. All in all, excellent work, and I'm looking forward to your next submission!
Spikestuff: Publishing.
Last Edited by adelikat on 11/5/2023 9:28 PM
Page History Latest diff List referrers