Placeable cursor on the screen, plus editable textboxes for its exact co-ordinates, plus progressively faded targets for the cursor locations of the past X frames?
That would be true, if not for the fact that the NES were also affected by external conditions; slight variations in timing crystals and so on.
It's been discussed already - you'd need not only the right starting state, and for all physical conditions to exactly match the 'averaged' assumptions of the emulator.
I actually thought the issue brought up a reverse problem for me: Should TASers be allowed to start with whatever uninitialized memory, entrophy, etc they want? It would still be a run possible on the actual console, just as likely as a run with any other starting memory and entrophy (however vanishingly likely that is; hmm, instead of making static TASes perhaps we need to make smart movie files that adjust their input slightly whenever the output of the console isn't exactly what's expected. ;) )
You guys should check up on your algorithmic information theory. A true random number is uncomputable, so no algorithm can create one (there do exist approximations, but they all deffer depending on your formal language).
If so, surely the universe cannot 'compute' a random number either?
One way to find glitches would be to experiment with things that are often poorly done in other games; for instance, you can zip in similar ways in both mega man 1 and 2, so if a new mega man game came out (10!) you might try and reproduce zipping in the same way.
http://slacktivist.typepad.com/slacktivist/ contains intelligent discussion on religious/evangelical issues. Its riffings of the Left Behind series were what got me hooked.
Another thing I do daily, if it counts, is watch youtube.com/user/raocow 's new videos. It is essentially a stream of consciousness on top of super mario world/yoshi's island hacks, so it's not for everyone.
When I read the title what came to mind was an image of puzzle games where there's no hidden information - picture a sudoku game for instance, where the only thing a TAS can do is fill it in really fast or in creative patterns.
Maybe instead of a "heavy luck manipulation" tag a run could have a luck manipulation quotient or something like that. It would indicate how probable it is for the rng (if not manipulated) to spit out the numbers needed to make the run possible. That doesn't include random events that don't influence the run's time. It should, however, include luck manipulations for entertainment purposes. For games where it isn't understood how the rng works, this value would have to be approximated.
The higher (or lower) that value and the shorter the movie, the more densely luck manipulated it is.
Just a suggestion, as I don't see people ever agreeing on what constitutes "heavy luck manipulation".
It would mean a lot of work, so my bet is that nobody will actually ever calculate these values. ;)
The problem is it's not just a function of how improbable the RNG's results are, but how muchg time would be lost if the RNG was not predictable.
Like, let's say there's a manuver you can do that has a 1/10 chance of happening but saves a second to do so. A TAS would always take advantage of it but a real time runner might abstain if the drawbacks are too bad. You can think of it like this: If the RNG was related with a perfect RNG that wasn't altered by the state of the game, how much slower would a tool assisted run (entering input frame by frame with perfect knowledge of the game) be compared to the manipulatable, predictable RNG?
Does that make sense?
Weird timing. I was just about to do another run of this game to see if I could manipulate Waldo to be right under the cursor every time. It would be an interesting technical achievement, though still not an interesting movie. At any rate, I suspect that this run is improvable... at least in terms of length. Not in terms of entertainment. In that regard, the game is beyond help.
It'd be worth doing for the sake of novelty, kind of like the king's bounty video but not as dramatic.
Shmups are trivial to TAS, but there's a shump bullet pattern emulator called YGS2000 which has a 'level' parameter, which scales almost every pattern up to ludicrous levels. (Some produce more interesting results than others.)