After playing VVVVVV, I started wondering whether a 3D equivalent would be worthwhile.
I soon shot that idea in the head with the note that one of the reasons 2D games work so well is because of the "division symmetry" in 2D worlds. The frontal dimension takes on an opposing role to the lateral dimension, and the vertical dimension is a vector with (possibly zero-valued) components in each. Both tunnels and walls are fundamentally one-dimensional (the other dimension is their "thickness"), and each surround the other.
In 3D, you can either have a one-dimensional tunnel surrounded by two-dimensional polar mass, or a one-dimensional wall surrounded by two-dimensional polar space. These two possibilities are dual to each other, and lead to two different types of games. The first case gives you a game where your movement is limited to a network of tunnels, which can give the player a claustrophobic feeling which is not felt from playing 2D games. The second case gives you a wide-open game, more akin to the real world, where the player is theoretically free to explore a vast area, and while this can certainly look very nice, it often makes players unsure where they are supposed to go next and makes developers' work much harder since they need to either spend a lot of time map-making for very little gain, or introduce invisible walls (or Insurmountable Waist-Height Fences) to unrealistically limit the player's movement. So both leave room for improvement.
In 4D, as well as extrapolating the cases of one-dimensional tunnel/wall surrounded by three-dimensional polar mass/space, there should theoretically be a new option, with two dimensions each, which should exhibit the same type of "symmetry" as 2D worlds. Both tunnels and walls would be two-dimensional polar, each surrounding the other. If you want to imagine it in 3D, first imagine a 2D polar space (as your tunnel/wall), add some 3D thickness to it, discard one of the first two dimensions to make it appear as an infinite cylinder, and then the mass/space occupies the remaining two dimensions around the cylinder. I have briefly touched on this note before.
If we were to make a 4D "Metroidvania" type game (which we could actually play on our 2D screens, and without making it something as simple as the various hypercube-cell maze games that've been popping up around here for a while now), I suspect the easiest way to do this would be to discard two dimensions at the interface (i.e. input control interpretation, and output video rendering), choosing which dimensions to discard based either on game scripting, user selection or a combination of both. For rendering, the discarded dimensions could factor into brightness/color and parallax/stereoscopic information. To help see both extra dimensions at once, the renderer could apply a rotation in the ZW plane (assuming WLOG that the screen is the XY plane) prior to discarding, based on some superfluous information such as the position of the player modulo some factor, or a periodic time counter, etc.
I imagine that in such a 4D game, duocylinders and toric surfaces would make natural appearances all over the place and we would probably have to come up with some sci-fi explanation for 2D polar gravity and oddly-shaped planets...
Any thoughts? Would this be at all worthwhile, or would it be better to stick with 2.5D with partially controllable (and partially scripted) 3D rotations?