Umihara Kawase as a series leaves me as a bitter, hopeless wreck more often than not. If you see me stampeding down the street swinging a fish hook and shouting "LOOK WHAT YOU DID TO ME", then get your medications checked for traces of severe hallucinogens. I mean, what?

And please read the rest of the submission text before you bring them up. There are two levels that are slower than the current IL TAS records, and there are reasons for that. Should any legitimate improvements be brought up, I will either cancel the submission or ask for it to be set to Delayed so I can implement them.

(removed, sorry!)

This run is 190 frames faster than the published run in terms of in-game frames and should this time ACTUALLY match or beat all but 2 of the current TAS records on the DS version. Why all but 2? Let me explain: Turns out that the physics between the PSX version and the DS version are tweaked ever so slightly in a way that is completely unnoticeable in realtime play. The updated videos from TKDL (who else) had input display: Direct frame-for-frame transfer of the input results in trajectories being ever-so-slightly off between the two versions. The DS TAS records were also done without skipping the intro flyovers.

Combining these two facts leads to the first two Fields being 2 frames and 1 frame slower than the DS records, respectively. I spent nearly all of my time working on this run working on just those two Fields to try to eke out every frame I could, to no avail. Both Fields are still improved over the published run (especially Field 0, which contains over half of the improvement by itself), but compared to the DS version runs I couldn't avoid losing those three frames.

On the other hand, Fields 15 and 35 are actually faster than the DS TAS records, by 3 frames and 1 frame respectively. So overall, I'm a frame faster than the DS records for this set of stages.

It's easier for me to work with, to put it simply. Call me when DSHawk happens and I'll think about coming back and giving the run one more go.

No Field comments this time, as there's not really much to say about them.

Somehow, after a seriously rough start on the first two Fields, I managed to power through the rest of the run with ease. I can't tell if it was just me getting way better at Umihara TASing or if I just got ridiculously lucky, but either way I'm glad to have it finished. That being said, this time I'm really not going to return to Umihara TASing... Not for a long time, at least.

Umihara TASing at this stage is just pure optimization. I don't see any new routes showing up for any of the stages anytime soon. Even if I were to do something like make a sightseeing run for this game, it wouldn't be any different than the usual process: Find the record, implement it and optimize it. I'm not a fan of doing that. It feels cheap, to say the least. I like finding out new things myself, redoing everything from the ground up to make it truly mine. This... really isn't mine. It's my input, my tweaking and optimizing, but I still don't feel like it's my run.

As I said earlier, the main thing that'll get me back into proper Umihara TASing is the implementation of DSHawk, or at the very least a more stable DeSmuMe. In the meantime, maybe I'll try to find some new ridiculous routes or something, so the run actually does feel like something I created instead of simply optimizing someone else's work.

...Or maybe, once 3DS/Vita TASing comes about, I'll be at the front of the pack for Sayonara Umihara Kawase. We'll see about that when we get there.

Might as well flesh this out more while I'm editing other things.

ars4326: Judging!

ars4326: Hello again, Samsara. Very impressive how you came back to this one in short order, combing back through and finding improvements on almost all of the stages. Comparing side-by-side, fields 00 and 35 were especially noticeable (with 00 being the most significant, obviously). All in all, outstanding job and determination! Perhaps you and TKDL should collaborate on a future Umihara run?

Accepting as an improvement to the published run!

fsvgm777: Processing.