Good work so far! I'm currently recovering from an injury, and can't do any TASing work, but I have been doing some research with my own physical copy of the game. A few things I've found:
1. This isn't related to the research, but figured I'd get it out of the way: When I first became interested in this game, well before my injury, I made a
WIP of my own and later dropped it and went back to lurking. Just based on a visual comparison, it seems to be saving frames over yours not counting the difference in filenames. Someone want to double check?
2. I'm fairly confident that the only difference between 100% (the stated goal) and any% is just doing the tutorial plus the silver and gold bonus levels on each stage. The fanfare at the end of each stage is the same length regardless of performance, and the only way to lose stars is to get a strike (which is slow, as you can't do anything except pause during its animation) and exceeding the initial time limit (which is
really slow, five minutes per puzzle at minimum). Just figured I'd check, though.
3. I have a theory as to how the mechanic I'll call "tunnelling" works. As you've seen, tapping each block individually to break it is slower than one might expect because it's throttled (unless maybe blocks are tapped in a zig-zag motion, so the block breaking animation doesn't get in the way? someone check that for me). However, if you do a long press on a block in break mode, you enter what I'll call "tunnelling mode", where blocks break pretty much as soon as you drag over them if they're in the row or column selected, and do nothing otherwise. You do not need to drag over all the blocks you want to break in any specific order, and you can drag over blocks not in the selected row or column without breaking them. The game selects a row or column apparently via a combination of the specific block on which the long press was done, and whichever dimension (width, height, or depth) most closely aligns with the direction pointing from the camera "into" the screen. (A RAM watch would be useful here, though I have no idea how to find the address in question.) This mode ends once the stylus is released from the screen. I suspect this will be very useful all over the place for breaking blocks faster than individual taps would, though one has to account for the time cost of moving the camera if needed to set this up and the long press in the first place. I suspect it's faster overall for any row or column where you need to break at least maybe four blocks, though this needs testing; I suggest Easy level 1-D (accessible from the start of the game) as its leftmost column is nine blocks tall and only the topmost is filled (i.e. must not be broken).
4. Finally, sometimes it might be necessary to use the slicers to access blocks in the interior of a puzzle; but it seems like it is slower to finish a puzzle when "sliced in" as the game pauses for a moment to slice out before playing the level end fanfare.
That's all I have for now; hopefully soon I'll be able to test some of those things in emulator and make a more complete WIP.
