Is it possible to program code into the game that alters models (like combining vertices and edges) so that certain models would have a lower poly count? I'm not sure what this would do to lag though. Having to watch every model/object's distance from the camera may end up taking as much memory as any other method.

Anyway, with code that would automatically cut triangles off the model, you could have high-res models up close, and then very far away there could be models with only 6 triangles, then just about every extreme in between.

Ex:

20m away: 500+ polys
100m away: 200-500 polys
250m away: 80-200 polys
350m away: 25-80 polys
500+m away: 25- polys

note: I'm just learning to program games like this, so if I saw something that is either really simple (like the LOD thing I said before) or something totally outrageous, please tell me. All I'm trying to do is think of concepts that maybe nobody else has thought of before, however unlikely that is to happen...

You can generate different details of polys through a Geometry shader
http://en.wikipedia.org/wiki/Geometry_shader

All of this isn't new, most games do them but it isn't exactly easy.

Ok then...

I learn something new everyday. For some reason, it never comes from my schooling -_-

So now I'm interested in how many triangles did that Star Destroyer have in the 10,000 VSDs pic? And, what would be the most detailed amount of triangles that around 3,000 models on screen could have? (including particle effects, textures, lighting and shading, and sounds)

Geometry shaders require extra processing time during rendering and newer graphics cards. I believe that the usual way of doing this is having multiple models stored in video memory and simply swapping them as needed. Also sprites are often used for models that are really far away (see Total War). These low detail models could be automatically generated at the start of the game, but this needs some decent algorithm.

And as for triangle nubers..imho LOD tries to hide invisible details. So that shapes or curves that are rendered into one pixel (or otherwise indistinguishable) are merged together, detail "popping" is undesired artifact after all.

That is what I've decided to do for Dominance. The EaW-style battles will use lower detail models while the Battlefront-style ones use higher detail. Then the far away objects will use low (very low) detail models. I haven't thought about using sprites though. But everything in Dominance is basically in constant motion so how might that work?



So, is LOD something I have to learn to code myself, or does the graphics card or the code do it for me? Sorry for the stupid question, I just really don't know..

Pretty sure LOD is handled by the graphics card. Once you set it up does it automattically... could be wrong though

No, graphics card knows nothing about higher level abstractions above data (like LOD, animations, etc.) These must be managed by the 3d engine.

Hm... unless you're changing the game in more ways than I think you should, won't the galaxy size itself, I mean the maintenance possible to derive from planets, make sure that there aren't thousands of capital ships in a battle?

EDIT: never mind me, I guess if you're designing a moddable game, anything can happen.