1. This idea is awesome.
2. Movies made with these goals probably won't be terribly interesting to the general public. We should decide what should get done with them - reasonable options might be: a) submit them anyway, with the knowledge that they will be gruefood; b) post them in the game's thread; c) make a subforum.
b) has the problem that people who don't watch that thread won't hear about them; c) has the problem that they will probably never be popular enough, and a) has the problem that it might feel like spam to people who don't care/they will still show up on the front page under 'recent submissions'. However, I like a) best anyway and think that's what people who make MIP movies should do.
3. I can't wait for when there's a movie that requires some serious trickery to figure out how to be allowed to continue to hold a necessary button until the last frame, thus vindicating Warp's position ^^
4. Actually, though, I think that counting button presses but not frames held is not as interesting as counting each button/frame. We're talking about minimal input, why should we be incentivizing holding extra unnecessary buttons?
5. It does kind of feel like 'holding right' is less input than tapping right every other frame though, and I think that's why the 'count button presses' metric has caught people's fancy. However, holding start all the way through a movie doesn't feel the same. That makes this kind of hard to formalize.
6. At the risk of making this enterprise utterly bizarre and obtuse looking from the outside (but who are we kidding, it already is) we could define two classes of button: those that dislike changing state, and those that dislike being in the 'pressed' state. Then, for each game, there would be 2^n possible movies to make, where n is the number of buttons on the controller, one for each possible set of assignments of a button-types to buttons.
7. Of course, such a proliferation of goals is terrible, so we want to consolidate them. Luckily, this is easy! We can easily define a comparison function between two movies of the same game made with different button-type assignments: simply evaluate the "unhappiness" of each button in each movie, and pick the one with the lowest total! Thus, choosing a button-type assignment set becomes part of the challenge of making the optimal run!
8. The preceding may have been confusing. Let me illustrate with an example. Let us say we have a simple game with only 2 buttons, A and B. Let us say that the game is always completed in 6 frames, if it is completed at all. Now let us consider two competing runs, one with this input:
Frm A B
1: 1 0
2: 1 1
3: 1 0
4: 1 1
5: 1 0
6: 1 1
and the second with this input:
Frm A B
1: 1 0
2: 1 1
3: 0 1
4: 0 0
5: 1 1
6: 0 1
The first movie will assign button A to dislike change and B to dislike being pressed. It has a total unhappiness of 5: 2 changes for A, and 3 frames down for B. The second will assign A to dislike being pressed, and either preference to B - both of it's presses come in groups of two frames, which are worth 2 unhappiness points under both preferences schemes. It's total will be 7, so movie 1 will be the victor.
It may at first seem kind of unfortunate that the crossover point is at an average of two frames per button activation, but given how most games work, I think it is reasonable. In many games, jumping and shooting only take one frame, and thus those buttons can be profitably assigned the 'dislikes being held down' preference.
9. I've chosen above to treat time before and after the movie as ... existing. And being real. By which I mean, you can't say you've been holding that button forever and will continue to hold it forever after the movie stops. My motivation behind this is that it incentives the thing that originally motivated this entire idea - not holding start the whole way through. If pressing start on frame 1 and holding it all the way through was 0 unhappiness points, but pressing it once on frame 1 and never again was 1, that would do the opposite. As it is, holding it all the way through is 2, which is what I want.
10. While making the example, I had a really silly and off-topic idea, and since you clearly haven't read enough words in this post yet, I thought I'd share. It is a challenge of a wildly impractical nature: I present you with two very different input files of equal unhappiness. You design a game to which they are both MIP movies, and which has consistent/reasonable input semantics than can be learned and played by a human.
... I might actually give that a shot with the two 6-frame inputs I used above. I'll prolly say the game runs at about 1 fps.
Five hundred yes votes.