Submission #4646: Alyosha, ars4326 & Samsara's SNES Umihara Kawase "Sightseeing" in 11:44.77

After over 10 years and somehow no one beating us to the punch, an improvement to the Umihara Kawase "sightseeing" run rappels its way onto the Workbench!

Samsara: ...No.

Samsara: NOW I've run out of puns.

Samsara: Umihara Kawase (海腹川背) is both the name of the game and the name of the player character. Umihara Kawase (the game) is a unique physics platformer (a "rappelling action game" if I'm reading that katakana right), released only in Japan at the tail end of 1994. Umihara Kawase (the character) is somehow 19 years old. I'm just as confused as you are. Armed with a fishing hook that can attach to nearly anything, Umihara-chan blazes through a bunch of Fields, possibly catching mutated sea creatures, though not too many in this case, as it's a lot slower to do so. She does, however, catch more than her fair share of walls, floors and ceilings, though I'd imagine mounting a wall on a wall would either be the easiest thing in the world or the hardest, depending on how you look at it. Let's not even go into gutting the wall and serving it for that night's dinner.

The game is the first of a still running series, believe it or not! Recently, Sayonara Umihara Kawase, the latest entry of the series, has come to North America in the form of Yumi's Odd Odyssey, localized and published by PRESS THE START. Excuse me, I mean Natsume. What better way to celebrate a game that's been out for a year than with this run!

Samsara: This run is 5351 frames faster than the currently published run by Bob Whoops. The currently published run is the longest non-obsoleted movie on the site, having been published in August of 2004. There are only two other runs that haven't been obsoleted since 2004. If anyone wants to keep the obsoletion train a-rollin', have fun with Combatribes!

The "sightseeing" branch is the closest the game can get to "all levels" in a single run. It plays about half of all the Fields in the game (26 levels), using the longest path you can take. There are actually 3 different paths you can take that all have 26 levels and only two bosses, and all three of them differ in the same section of 5 levels. Bob Whoops' published run actually takes the longest of these three paths, so in the interest of saving as much time as possible we decided to take the shortest path. The path change saves about 15 seconds by itself. Even with the path change, there are still 21 common Fields between our run and the published run. All of these Fields were improved, most of them by over a second and quite a few by several seconds.

The re-record count is an estimate: We primarily used TAStudio to create the run, working off of Bob Whoops' run as a base. Throughout all the testing on multiple files (I've got 20gb of .TASPROJ files archived just in case), passing them back and forth, checking and re-checking each others' work, resyncing each Field after improving prior ones, all the times we've gone back through the game due to the various improvements and path changes, and dealing with TAStudio still being in beta, I honestly think it's an accurate number. This game does not play kindly in regards to syncing: Improvements in earlier stages often leads to different lag in later stages, meaning all the input has to be changed around for every stage past the one we improved, and occasionally we had to wrack our brains to find improvements in areas that just straight up wouldn't resync at all. Between the three of us, we've gone through and redone the entire movie between 5-10 times for various reasons.

Initially, we had gotten very close to completing the movie by ourselves. We'd finished every normal stage except for the crab battle, we'd started writing up this submission text, I would've whipped up a temp encode and it would've been on the workbench... However, out of curiosity, I started looking through Umihara videos on Youtube and stumbled across another TAS, made by the Japanese runner TKDL. They combined every Field and exit into a single video, and more importantly they had a few strategies that were definitely faster than what we came up with, so after a brief moment of angst from yours truly we decided to give our run one final pass to implement and optimize these changes. Despite the similar strategies, each Field was redone by hand with our own input, as we only had individual level videos (for most stages) to go off of. We do not have a direct frame comparison for this reason. The individual level videos do have frame counters, but seeing as they were all made in Snes9x 1.43 on a modified ROM it would still not be an accurate comparison. After one last pass through the game, we either match or beat all of TKDL's times. This run is about 3 seconds faster than a comparable run would be using TKDL’s exact routes re-synced to bsnes.

For anyone wondering why we didn't use the DS port for reduced loading times and the more accurate timer: There seems to be a slight change to the way picking up enemies works. Namely, the radius is a bit larger. There are a few strategies we use in this run that wouldn't work in the DS port due to us being a single frame away from catching enemies in certain cases. The line can't be cast while Umihara's putting something into her bag, and it takes about a second or so for her to do that. Adding a few seconds of gameplay to reduce loading times really doesn't seem like a fair trade-off.

...Also, DeSMuME doesn't have non-linear TASing yet. So there's that.

Samsara: Forgive the odd layout of the table. I wanted to make the path change clear while detailing how much time it saves, and this is the best way I found. If only I could optimize my submission text the same way I was optimizing this movie!

Alyosha: So where do the extra 185 frames come from? It turns out that if you compare input lengths per level, instead of using the frame timer, that actually about 10 extra frames per level would be saved, indicating just how big a difference exists between bsnes and snes9x. Working out the calculation this way shows we actually lost time time due to the emulation differences, as would be expected.

Samsara: These Fields were redone directly from flagitious' any% run, and contain no improvements of our own. We tried, believe me, but flagitious did some insane work here.

Samsara: I admit that I wasn't too much involved with most of the earlier Fields. The most I did with this Field was just optimizing the ending by 5 frames. I know, I'm amazing. You don't have to tell me.

Samsara: This Field went through several iterations and lots of improvements. It's one of the stages that I think really proves the benefit of working with multiple people. We were initially following Bob's route a little more closely until ars came up with the route change after the jump at the beginning. After some tweaking, we had more than our fair share of snail troubles that kept us thinking for a couple days. After some hard optimization, we came up with the strategy you're seeing here. Or rather, that you'll be seeing in the run when you watch it. Or that you've already seen in the run if you've watched it. Look, the point is that you watched it. Vote Yes! Like and subscribe! Follow us on Twitter! Reblog our Tumblr selfies!

Samsara: See my Field 02 comments, except I don't think I even managed to optimize this level at all. The part about me being amazing still stands, though. Really, this run should have been called Samsara Kawase.

Samsara: Hook, line and sleeker. I've been saving that one. No, I didn't work on this Field. How did you know?

Samsara: *whistles innocently to distract you from the fact that I had nothing to do with this Field either*

Samsara: Hang on, I think I did something here. I did. Okay. That strategy at the beginning looks a little awkward with the way Umihara hits the platform, but it's intentional to make sure she stops right there instead of flying past it.

Samsara: I improved exiting the Field by a bunch of frames, using TKDL's strategy as a base and improving it. Now let me just casually step aside and let ars talk about this awesome boss battle.

ars4326: Going into this one, I volunteered to be in charge of scripting the two boss battles and strove to create the most entertaining TAS choreography possible. For this tadpole boss fight, my inspiration here mostly drew upon the question of “How would a little girl (who’s strangely 19, but whatever) react against a giant, goofy tadpole?”

Starting out, I relied more on making a narrative over showing off multiple, TAS-related tricks: Umihara-chan begins by confronting the obesely overgrown tadpole and discovers she...can’t quite hook him in like the others. Exasperated, she turns and spins around again, and he’s suddenly gone! She turns again the other way, where did he go? And, at one point, after a brief chase, Umihara even corners her gigantic foe and plants a little kiss on the tip of its nose (before being relentlessly pursued again). It’s little storytelling elements like this that I believe, when properly utilized, really add to a presentation.

Anyhow, the tadpole is scripted to make six complete passes, back and forth, across the platform above the pond. From the second pass onwards, it lays increasing numbers of tadpole eggs which then hatch and participate in the battle. Plenty of creative ideas resulted from racking my brain and jump-starting my imagination, such as sneaking up behind, a la Metal Gear Solid, and attempting to “steal” some eggs. Or, precision hook-lining underneath the platform to skillfully catch a swimming mini-tadpole, just as Umihara lands on the edge of the island off in the bottom-right corner. I even utilized a clever trick from Bob Whoop’s run, involving ricocheting off of enemies to prevent a stunned Umihara-chan from sinking down into the pond.

Overall, I believe these boss fights are a remarkable addition to the tight optimizations of the field stages. Hope you all enjoy!

Samsara: From here up until the end of the movie is where I can actually start talking about the Fields! This is around the point where we hit our real groove in the initial run, improving it by a fair amount even before incorporating and optimizing the new strategy from TKDL's run. The ending is pretty precise.

Samsara: A number of new strategies on this stage lead up to a solid improvement. I had a running theme of "Optimizing Endings" in this movie, where every time the file was passed back over to me, I'd almost always find a way to improve the ending of each Field by at least a few frames. This one's quite a bit more significant, saving a full second even over TKDL.

Samsara: This is a stage that got passed back and forth a few times, each time getting a new strategy change. Alyosha's result here is a lot more fluid than my initial attempts. And to think I said I could actually start talking about the Fields in this section. I think I optimized climbing the ladder a little bit. Man, I shouldn't be writing submission texts at 2 in the morning.

Samsara: Whoa hey, where are you going, Umihara-chan? That's not where Bob went! Holy hell, this is like the weirdest desync since Wario Land! And it's exactly the same because it's intentional. This is where the route changes from Bob's run: The next 5 stages are all very much shorter than the stages in Bob's path, far more than we could ever optimize out of them ourselves, so look forward to that! Unless you've already watched it. And voted Yes. And liked and subsc-

Samsara: Alright, so a lot went into this stage, an- Oh, it's over. How 'bout that.

Samsara: A lot of my comments in this submission text before rewriting were "We optimized Bob's route and strategies!" Just wanted to let you know that. Anyway, we optimized TKDL's route and strategies!

Samsara: The introduction and sad farewell of the Climb Through Floors glitch! This has been documented on the Game Resources page, but until now hasn't been used in a submitted run. Unless it was used in a submitted run. Then you're free to make fun of me as much as you want. Actually, don't. I bite.

Samsara: Swing swing. I came up with a new strat! That's it! Hooray! *blows up a balloon and lets it deflate all around the room*

Samsara: This is my favorite Field in the new run. It contains an interesting little glitch at the end, plus a lot of jumping! This Field probably contains the most jumping out of any Field in the game. Isn't that exciting?

Samsara: Alright, our little path detour is over now. I swear I tried this strat back during the initial optimization session on our original movie, but it didn't work out. I guess it actually does!

Samsara: I'm getting a headache just thinking about this level.

Samsara: 3 + 1 = 4, and 4 is death, which is pretty much how I felt while working on this Field. I must've redone and re-optimized this stage about five times, after finding new strategies and accidentally saving over them while testing different strategies. I'm not very smart, if it isn't obvious already after my long posting career. For the record, that sliding panel to the right of the exit door does not open from the right side, which makes me question why it's even there in the first place. It could very well have just been a wall, but nah. Umihara Kawase is a fickle mistress sometimes. The game, not the character.

Samsara: Lots of strategy optimization here, and it continues with the theme I just made up of Umihara Kawase (the character, not the game) squeezing through really tight holes. Is there a metaphor here? You be the judge!

Samsara: There's a phenomenon I've talked about before when it comes to working on improvements. Quoting myself: "...when I've looked up information before starting a run, I overlooked things that I should have tested, assuming that the information laid out before me is all I needed to know." Going into this stage for the first time, I hadn't even considered trying to go through that hole. Bob didn't do it, so I assumed it was either impossible or slower, and given how recently this route change was added, I can only assume that all three of us felt the same way. However, Alyosha found that it was not only possible, but faster by several hundred frames.

Not to get all preachy about TASing outside of another long-winded ramble I'm writing, but I think this improvement really justifies two of my philosophies: "There's always a faster way, I just need to find it" and "My knowledge is just as important as anyone else's". As I said, this was a very recent improvement, having only been added within the last three weeks, and until then it was overlooked. We had even optimized Bob's initial route a couple times before trying this way out. If you're new to TASing, or even if you have some experience, take this to heart: Test everything, no matter what. Never assume that what you did or what anyone else did is "perfect", otherwise you'll overlook the simple things.

But at the same time, use your resources well. There's an entire world of TASers out there that you might not even know about. This isn't the only scene. Compile your resources and check your list twice, but remember that your own knowledge is just as important as what you may learn from other people. Our strategies and TKDL's strategies ended up being largely similar, with them being faster in some places and us being faster in other places. Had we not heavily optimized Bob's original work ourselves before finding TKDL's run, I don't think our final submission would be as optimized as it is now. It gave us a chance to really learn the game's physics on our own, and once we got new angles to consider it was a lot easier to find those tiny optimizations all throughout.

Anyway, enough preaching. As for the stage itself, once the new route change was implemented, it was further optimized, bringing the longest non-boss stage in the previous run down to one of the shorter stages in this run. I really like how everything turned out here.

Samsara: The saddest part of this new route is that I can't use the best joke I made in the last submission text, because we jump right over the octopus instead of using it to boost ourselves up. For more information on how sad I am over losing the sick octopus cancel, please check out my cephalopodcast. THANK YOU! *drops mic*

Samsara: Four minutes of awesome!

ars4326: After the tadpole fight, I took a look at the crab battle and realized it was about twice as long as the previous one. So, like my last semester in college on the verge of finals week, I knew I had to “step my game up” and work toward making this fight even better. Especially after observing how hard my teammates Samsara and Alyosha were working, I wanted to honestly look at this one coming out of it and say “That was my best work”.

I hope it shows!

Similar to the tadpole, the crab battle consists of six ‘waves’ or passes where the boss methodically moves back-and-forth across the screen. There’s also six conspicuous wooden platforms above, one of which is blocking the exit. The boss’s pattern runs as follows: Beginning from the right, the crab stalks toward your current position (wherever you may be) and, upon reaching you, stops to emit a series of bubbles; some of which may contain little crabs. Afterward, the boss continues leftward and briefly disappears off-screen. Upon returning, Mr. Crab (going from left to right) approaches underneath a wooden platform and dramatically shatters it with one of its claws, before continuing rightward and off-screen again. This pattern continues until the last platform (blocking the exit) is shattered.

My game plan going into this was to do something creatively different on each separate ‘wave’; whether it was precision-timed hook shots popping all of the crab’s bubbles, or Matrix-style bubble dodging via a single swinging sequence directly next to the crab. I’ll spare the redundant details and just state here that I’m very satisfied with the end result. If anyone has any specific questions, though, I’ll be more than happy to answer them in the thread.

Samsara: This was basically all Alyosha, and I feel weird writing this while ars is talking about how hard I was working directly above me. Collaborative writing is weeeeeird.

Samsara: This is always a really hard section to write, but for this game... It's nearly impossible to list potential improvements. The game is so open-ended, there are so many ways to tackle each Field and so many ways you can use the fishing line that it's hard to narrow down improvements to anything specific. Are there still improvements left? I'm sure there are - there are improvements in every published run on the site after all - but I can't tell you what they are or even where to begin looking for them.

Samsara: I'd like to start with a positive note. I've always kinda felt bad about my participation in this run, given that said participation was basically nada until mid-March. We, as a team, had intended to do the project together from beginning to end, but a few things came up in my life that left me inactive and unmotivated for pretty much the first month of the project. I left before we really even started working on the run, and came back in a few days after Alyosha posted the last WIP in the thread. I then proceeded to find tons of improvements that I really should have been finding this entire time, and capped it all off by finding TKDLa's individual levels on YouTube... literally the day before we would have finished and likely submitted the run. I don't really think I apologized enough in private to my teammates, so I'm doing it publicly: I'm sorry.

Wait, that wasn't a positive note at all. Oh well!

That being said, once I got back into the swing of things, the run was a blast to work on. I've always really liked this game ever since I found out about it through TASVideos, and I'd always wanted to try my hand at improving this run in particular, though for the longest time I hadn't considered my skills up to par. DTC5 was sort of an eye-opener for me in a few ways: It taught me that I can put a hell of a lot of effort into something if I really set my mind to it, and it taught me that other people are an incredible boon to the TASing process. Alyosha and ars are amazing teammates: They're both highly dedicated workers who have an incredible knack for finding things, and they both have attitudes far more positive than mine will ever be. I still consider most of the run to be their work.

All in all, I'm glad all the work is done. I remember back before we started the project, when we were talking about it, and one of my first offhand comments was “I wouldn’t be surprised if we could save a minute” or something to that effect, and now we’re sitting here with a full minute and a half improvement and I’m honestly looking back in mild shock. The fact that we’ve come this far is amazing. Honestly, it's been a hell of a ride! ...More so for ars and Alyosha than me, I assume, but still! Maybe you'll see another run from the three of us in the future, who knows?

Samsara: I suppose first and foremost, I'd like to thank Bob Whoops, flagitious, and TKDL. Bob's currently published run and flagitious' any% run were two of the first runs I watched on the site back when I first found it in 2006, and they've always been runs I've held in high regard. I kept telling myself "You should improve the sightseeing run!" for the longest time, but always felt like I didn't have the skill to power through it, and now that this run's on the Workbench, I know for a fact that I still don't have the skill to power through it. TKDL's more optimized TASes - though I'm still a little mad about finding them so late - were also a huge benefit to us, not just to double check our own new Field routes and strategies, but to double check the path through the entire game. We would've kept with "Bob's path" otherwise.

Now for the personal ones. I gotta thank Alyosha and ars4326 without any hesitation, as like I said I still consider the run to be theirs. Even though I wrote the hell out of this submission text (depending on how much they add to it in post) and did some serious optimizing of my own, they're still the ones that put in all the work in the first place. They're awesome, and I hope to work with them again in the future. And of course, this wouldn't be a sappy personal thanks section if I didn't thank *grabs a soapbox, drags it to the center of town, stands on it, pulls out a megaphone and shouts* MY GIRLFRIEND *awkwardly climbs down off the soapbox, throws the megaphone off to the side* for supporting me doing this weird hobby she only kind of understands. She's awesome, and I hope to work with her again in the future.

...That's not what I meant.

Alyosha: The three of us make quite a strong team. Samsara’s eye for optimization and Ars’s eye for entertainment really made this run what it is. Go Team!

Samsara: [dead links removed]

Tompa Time to judge.
Tompa Great run, great response, great! Accepted as an improvement to the "sightseeing" run.

fsvgm777: Processing.