RTA timing ends on Frame 1630, the last input that enters the final room (as per the speedrun.com rules), although I'm not sure if this is appropriate for TASVideos timing and if the final room can be treated as an epilogue section or not.
ENCODE/PLAYBACK INSTRUCTIONS: The same caveats largely apply as my K.O.L.M. submission, although this one actually runs way way better.
The recording WILL play and display on-screen with the release version of libTAS stated above, and can be verified that way, but it will NOT encode- the console will spam errors, and the output video will be totally broken.
The recording will encode on an interim version of libTAS, with "Interim 9204a7d6 (2023-09-14)" (artifact link) being used to produce the temporary encode above.
Required arguments: "-g gl --no-gui". Vulkan does NOT work, even for encode; software rendering can be on or off, doesn't matter.
The Armor Games sound plays twice at the beginning.
The final room has rendering issues, but nothing that affects sync or gameplay in any fashion whatsoever.
DESCRIPTION: Good god this game is broken. Only one new tech is used as compared to the original K.O.L.M., pause abuse; it is used at the beginning of the run to skip mandatory dialogue and take control of ROB earlier than otherwise possible, and to be able to get through sun areas where you would otherwise instantly die. Decreased time over RTA comes from superior execution, such as not being distracted/confused by the visual noise from the pause spam, pause spamming as little and as briefly as necessary, etc. This took me roughly two hours to record, discounting the time I spent pissing around or responding to Discord messages and whatnot.
I didn't manually check optimization in #8609: Technoturnovers's Flash K.O.L.M. in 07:08.50 because it looked quick and was visibly faster than the human record. But after #8601: Technoturnovers's Flash Ball Revamped IV: Amplitude in 10:34.15 I decided to take a closer look at this submission, and every room is improvable by working with the game's physics a bit better.
Wiki: GameResources/CommonTricks#JumpingOffLedges describes one example of movement when a game has acceleration, but the same applies to bumping corners in other directions: it slows you down and you have to accelerate again.
Also with double and triple jump, depending on which direction you're moving, you often need to find an optimal frame when to do a midair jump, because after a certain height you start falling. And even if you avoid downward speed, variable upward speed at different points in the jump means you need to keep the highest overall speed across all 3 jumps, just enough to get you on a platform (if you need to jump again ASAP), or around it (if you need to go down ASAP).
I haven't studied the jumping arc in this game (you often need to watch memory for that, or in case of a Flash game, read its code), but let's imagine it's coded like this:
and you need to jump on a high platform, to jump up from it again ASAP. For regular play it feels natural to do regular jumps, and if their height is not enough, do double and triple jumps. But in a speedrun, you end up having too many slow periods in your jumps, because they get slower after the first 3 frames (in our imaginary case), and even slower after the next 4 frames. For that reason, for all the multiple jumps you have to do, you need to test if you can get on the needed platform by only using the fastest segment of each jump (3 pixels per frame), meaning you hit the jump button every 3 frames. If that's not enough, start adding the second fastest segments to each jump until you've exhausted both speeds (3ppf and 2 ppf), hitting the jump button every 7 frames. Only then start adding the 8th frame to each jump (1ppf), and so on.
I haven't done this in my test, because it takes time to figure things out and to test them, while I only needed to try my most basic thoughts on each level, to see how easy it is to improve.
User movie #638328711837129400Link to video
I ended up saving 00:04.08 seconds, but the jumps are absolutely still improvable. Since most of the game consists of jumping and running, I think the opportunities missed in this submission are significant, and I'd like to see a better movie instead. Ideally one that's faster than mine, even if it's only a couple frames saved in each room.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Okay, damn. The issues with Ball are obvious in retrospect, but I had zero idea about this since it's totally irrelevant to RTA speedrunning- I'm just going to get this over with and cancel it until I incorporate what you're mentioning. Thanks for the input
I can record my libtas workflow if you want, because there are useful things that may not be obvious to newcomers. To be efficient you need to know what you want and how to get it without too much overwhelming routine.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
I can record my libtas workflow if you want, because there are useful things that may not be obvious to newcomers. To be efficient you need to know what you want and how to get it without too much overwhelming routine.
Nah, I'm good- what happened here is that I just totally discounted acceleration as a factor in this game, because it mostly isn't a factor along the horizontal, and even isn't a factor in short single upward jumps for the most part, so long as there's enough room ahead of the target platform. Faster acceleration from double-jumping for single jumps or mini-hopping off of platforms just never occurred to me, and even if it did, I would think it would be in the "shaving frames" category rather than saving whole seconds, especially since I had already incorporated recoil-boosted falling from shooting upwards into the run everywhere. Thankfully, this run is teeny tiny, so I should be able to fix this along with the Ball recordings without much issue.
and you need to jump on a high platform, to jump up from it again ASAP. For regular play it feels natural to do regular jumps, and if their height is not enough, do double and triple jumps. But in a speedrun, you end up having too many slow periods in your jumps, because they get slower after the first 3 frames (in our imaginary case), and even slower after the next 4 frames. For that reason, for all the multiple jumps you have to do, you need to test if you can get on the needed platform by only using the fastest segment of each jump (3 pixels per frame), meaning you hit the jump button every 3 frames. If that's not enough, start adding the second fastest segments to each jump until you've exhausted both speeds (3ppf and 2 ppf), hitting the jump button every 7 frames. Only then start adding the 8th frame to each jump (1ppf), and so on.
I haven't done this in my test, because it takes time to figure things out and to test them, while I only needed to try my most basic thoughts on each level, to see how easy it is to improve.
User movie #638328711837129400
Link to video
I ended up saving 00:04.08 seconds, but the jumps are absolutely still improvable. Since most of the game consists of jumping and running, I think the opportunities missed in this submission are significant, and I'd like to see a better movie instead. Ideally one that's faster than mine, even if it's only a couple frames saved in each room.
