Review:
My general impression after watching this TAS was positive. Even though each microgame (probably) has no time save because each has a fixed timer, the author Pellets finished them in a variety of interesting ways.
The bosses also seemed better than human play and decently optimized, although there's definitely future possible improvement.
For due diligence, I compared this TAS section-by-section to the current human WR (using
a script I quickly made). Conveniently, the current WR is on emulator, so I was able to compare GBA frame counts directly.
The sections below are listed in TAS order (the WR played some sections in a different order) and use the names from
here:
Intro Games, menu: -0.150 (TAS: 4.152, WR: 4.302)
Intro Games, games: 0.000 (TAS: 1:06.619, WR: 1:06.619)
Intro Games, boss: 0.000 (TAS: 14.047, WR: 14.047)
Sports, menu: -0.904 (TAS: 9.694, WR: 10.598)
Sports, games: 0.000 (TAS: 1:37.375, WR: 1:37.375)
Sports, boss: -2.595 (TAS: 12.724, WR: 15.319)
Sci-Fi, menu: -1.188 (TAS: 10.179, WR: 11.368)
Sci-Fi, games: 0.000 (TAS: 2:42.002, WR: 2:42.002)
Sci-Fi, boss: -2.327 (TAS: 59.185, WR: 1:01.512)
Strange, menu: -0.351 (TAS: 9.710, WR: 10.062)
Strange, games: 0.000 (TAS: 1:33.809, WR: 1:33.809)
Strange, boss: -0.418 (TAS: 11.937, WR: 12.356)
Nintendo Classics, menu: -1.473 (TAS: 9.375, WR: 10.849)
Nintendo Classics, games: 0.000 (TAS: 1:37.442, WR: 1:37.442)
Nintendo Classics, boss: 0.000 (TAS: 29.266, WR: 29.266)
Jimmy's Remix #1, menu: -1.339 (TAS: 9.945, WR: 11.284)
Jimmy's Remix #1, games: 0.000 (TAS: 2:02.690, WR: 2:02.690)
Jimmy's Remix #1, boss: -0.853 (TAS: 14.884, WR: 15.738)
IQ, menu: -1.456 (TAS: 10.380, WR: 11.837)
IQ, games: 0.000 (TAS: 2:18.495, WR: 2:18.495)
IQ, boss: -2.159 (TAS: 31.141, WR: 33.301)
Reality, menu: -0.468 (TAS: 10.263, WR: 10.732)
Reality, games: 0.000 (TAS: 1:28.100, WR: 1:28.100)
Reality, boss: -3.448 (TAS: 39.881, WR: 43.330)
Nature, menu: -1.406 (TAS: 10.346, WR: 11.753)
Nature, games: 0.000 (TAS: 2:39.959, WR: 2:39.959)
Nature, boss: -1.607 (TAS: 38.960, WR: 40.567)
Jimmy's Remix #2, menu: -1.691 (TAS: 10.464, WR: 12.155)
Jimmy's Remix #2, games: -5.859 (TAS: 2:14.176, WR: 2:20.035)
Jimmy's Remix #2, boss: -2.159 (TAS: 16.173, WR: 18.333)
Anything Goes, menu: -0.853 (TAS: 10.062, WR: 10.916)
Anything Goes, games: 0.000 (TAS: 2:20.923, WR: 2:20.923)
Anything Goes, boss: -1.238 (TAS: 1:09.515, WR: 1:10.754)
[Total]: -33.954 (= -33.954 +0.000) (TAS: 29:03.886, WR: 29:37.841)
menu: -11.284
games: -5.859
boss: -16.809
Here, menus encompass everything prior to and after each of the 11 sections. Games are everything except the boss game.
The 5.859s of time saved in the Jimmy's Remix #2 games seems to confirm it can have a random amount of IQ minigames (perhaps it will always have a minimum amount, though?)
(For good measure, I also compared to the WR using the video times directly. That comparison generally has more noise because of time delta granularity differences due to the GBA frame rate. The WR also loses an additional 0.1 seconds during Sci-Fi because the emulator application itself lagged.)
So this comparison shows the TAS is faster or equal in every section. Technically, the TAS probably lost a bit of time in some parts if each boss timing was further subdivided, but it's not significant, as each TAS boss was net faster than the corresponding WR boss. (Getting even further in the weeds, "faster or equal in every section" is an imperfect metric, because a future TAS with more RNG manipulation would probably lose minor time to this TAS in some parts to save time overall.)
All-in-all, this TAS seems like it would make a solid first publication—though I have some potential bias: see below.
An improvement:
I went a bit further than necessary for review purposes and tinkered with the inputs.
This led to
a 7-frame improvement in the final "Run, dummy!" section by doing first-frame A-presses instead an autofire on-off pattern.
I also found a few other improvements to other bosses' RNG that would require redoing the entire TAS or major portions to sync (and probably require more tools and scripts to optimize well).