:D It has become a competition :P This was actually done save for the point projection code before I had finished my Realm of Battle scenario. Still can't believe I missed such a key part...

Oh, and one question about the source. What do angY and angX actually represent? Are they the angle from the camera to the object? Or are they the rotation of the points?

Ah! Your source shows me where I went wrong with my 'quick render' method in my own 3D render engine. With this, I should be able to finish my own by the end of the week. Interested to see how you will handle additional vertex points and faces.

I'm not sure that that's exactly how it works...In my Realm of Battle, the original Fog Of War was ~2200 objects, none of which did anything. It slowed the game down immensely, even when no objects were interacting with them! I believe that at about 3000 objects, the game will perform at the middle speed setting even if it is set higher, regardless of the code content of the objects.

I ran into a bug--about half-way through, the computer wouldn't move. I think it despaired of my superior gaming skills :P If you want the AI to do well, just make sure that it tries to get the corners at all costs, and keeps you from getting them. That is how you make the AI win.

Mouse-look=the code that allows the mouse to change the rotation. It's something I learned making FPS games with Blender, actually. And I did actually write the code for a rendering engine in Greenfoot a while back. I'll see if I can find it, but it is extremely, painfully slow. Something to do with Greenfoot not being able to solve 8000+ simultaneous linear equations every act() very quickly :P But I will definitely look for my old code or re-write for the benefit of those who want "true 3D"

...Not to be a nay-sayer, but this really isn't that different from the Castle Storm scenario. Technically, this is only 2.5D, since there is no width to anything. From what I've done on this, true 3D is possible, but I do not think Greenfoot/Java is fast enough to do anything with it. The best looking approach though seems to be mutantleg's Terrain scenario. Still, nice way to improve on the other methods!