My opinion:
1) There is no reason whatsoever to obsolete 1-item run, even though it takes only one item less than this. 1-item run is technically a low% and has value in itself, since it shows Kraid, Ridley and the entire Tourian beaten under "bare bones" condition while taking the shortest routes possible, unlike 100%. They don't share any major tricks, except the wrong room door glitch (which, however, wasn't used in the current 1-item run since it wasn't available back then).
2) Even if people decide to obsolete an additional Metroid run, it should be the "no bosses" run, since it has far more repetition (item hunting, refilling in Tourian due to lack of missiles), shows no new tricks compared to the other runs including this, and is in need of improvement, anyway.
As I've already stated, the argument is moot because it involves arbitrary judgement criterion (in this case, plain arithmetic; ie. "one less" or "two instead of one"), instead of more-or-less publicly accepted ones: the lowest possible amount of items collected (low%) and beating all bosses showing the natural fastest route (not possible to showcase in 100% run due to item collection detours).
The particular reason I dislike JFG is that it's a cheap and laggy ripoff of MDK, which is superior to JFG in pretty much every respect.
MDK would also make an excellent TAS with its speed abuse — too bad we have to wait quite some time until it's possible to emulate mouse movements without problems.
It's easy to slap together a run that is 3-4 seconds away from the best run provided you use enough rerecords, but the thing is that no-one expects such level of precision in a 3D game, anyway. Saving a frame in SM64 would be thousands times easier than saving a frame in SMB, and you're free to try.
I play much better now, I just need training. I only speedrun, like, once a month or so, so I screw up much more than you, but I already can do many difficult maneuvers more freely now. Training should cure the moments of indecisiveness and those screw ups. I feel I can do much better than that, sub-50 should be possible with more practice.
So, I've just beat my old single-segment record (00:59) on a new Snes9x 1.51. The new record is 00:57, and it's still full of suck and fail. If you want to see how SM looks and sounds on a new emulator, here's the movie file. (Will be hosted there until Microstorage starts accepting new SMVs.)
The main reason I've brought it up, though, is to demonstrate the differences in emulation.
Graphics: as you see, there's no more annoying line of pixels under the status bar, and that's perfect. The colors are a tiny bit more clean and natural (unless it's my imagination), and resemble ZSNES somewhat. Look especially close at some effects like Phantoon's death. It's MUCH better now.
Sound: I didn't like the default sound, and never will: 32 KHz + interpolation = mufflefest. And there are still some sound quirks noticeable in several places, where the pitch of a sound is different, or when one of the channels sounds quiteter, if at all. The amount of these quirks is similar to that of 1.43's, they're just a bit different.
Timing, etc.: I'm not sure why, but it feels as if there's more lag. There, however, isn't. Must be something with its intesity in certain spots (timed on powerbombs). Draygon's turrets attack faster, Ridley spews a lot of fireballs. Space/time beam doesn't work at all.
It isn't. Saturn will likely have to redo it once again, since it turns out that I gained 23-24 frames on it in the segment shown here.
Obviously, it isn't the limit (see the notes page), so he can improve my WIP by a few more frames if he does his best. :) Besides, he is much more experienced than me when it comes to figuring out the paths in complex rooms, so the real feats will come in the more distant future.
Thanks, those two rooms are the only ones I'm fully satisfied with, [almost] all the others need both speed and style improvements. If you find a speed improvement, feel free to edit the page and add it there.
Yeah, that was played in single segment, I'm that good!
Seriously, though, it's a side-effect of savestateifying it. The actual rerecord count is 5400 (relatively low, but I barely spent any rerecords on Ceres).
See above.