In particular, you should search for information, tricks, timesavers, routes, do test run semi-optimized TASes etc, until you feel comfortable enough to do the TAS itself, otherwise you'll have to figure out so much stuff all at once that you'll progress slowly and be demotivated.
Apparently Gohdan's hands can be hurt by bombs. It might be a good TAS strat to lay bombs in his hands while bombing the head, so they go off after he re-activates. It even fully takes out the hand, maybe Gohdan can be done REALLY fast now?
Link to video
It is possible to tell if someone used savestates or not if they are sloppy. Here are some ideas:
1) Making really risky, unnatural dodges.
2) Reacting to things before they happen ed: or consistently faster than human reaction speed.
3) "Jittering", in the case of -loading a savestate made while moving then proceeding to not move, or move in a different direction -loading a savestate made while not moving then proceeding to move. If this happens a lot you can leave 'clutter' in the timeline that's meaningless, but left in because it didn't lead to you dying and because you didn't rewind far back enough.
This is unlikely because it implies that some dev has to code the operation 'transfer memory card to another instance of dolphin running in parallel' and have it be sync-stable, TASable, etc. Given how often anyone would want to use this, is it worth not just the development time but the development burden (has to not break or break things in the future)?
Surely if a TAS is obsoleted by a real time run it'll be because of some new glitch that no one has got around to making a new TAS with yet, not because the TAS got out-optimized?
Couldn't you get faster times per track by completing more tracks outside of test drive, levelling up your test drive vehicle further, but yet this would waste real time?
Very nice TAS! I loved the metal gear solid world 'die and yet save' trick in particular. Lots of deathtraps breezed through like they didn't exist.
I think for publication purposes, at the very least the sound glitch in the megaman battle would have to be fixed (1 frame saving is not worth it) :)
As for updating the TAS before that to include 1 frame shooting, though, that's up to you - the bosses are already so fast, so it wouldn't be a large entertainment change.
Do you have any videos of this? The only glitch-into-walls I know off the top of my head are the one in Infinite Corridor and the one in late game Chamber of Extinction from the statue dude.
Looks good! I have a question, from a TAS perspective what are the mechanisms by which you manipulate enemy movement? You said that the RNG is popped whenever an enemy has to decide what action to take, but is that also influenced itself by where you are on the level / what you kill and when?
Do you know why physics being different could possibly be caused by emulator problems? For example, in NES games like mega man sprites are moved around by specific pixel and subpixel amounts, so for that to be wrong it would have to not emulate basic math (and then nothing would work). What's the difference between that and Mupen?
IMO the best way to get a stronger RNG for the next pokemon game is to flagrantly and publicly abuse it as much as possible. Nip the problem in the bud so to speak :)
(And the fact that programs like RNG Reporter exist mean that understanding of how RNG works is strong and that the capability is there anyway)
You don't have to play both players simultaneously, there's a concept called multi-track recording where you record one player, then over the same part of the replay record the other player.
My suggestion is that while you can't see boss hit points, you can see how hard you hit them, so you can use memory search such that every time you hit the boss, you search for all values that are (that amount of damage) lower than before. If it's not possible with the emulator's memory search functionality it should be possible with the program Memory Hacking Software.