There are two separate issues here IMHO.
First, if you’re working for games you will almost certainly use normal maps, or if it’s an even less capable platform then you’ll rely on pre-painted shading in the textures. In both cases the focus should be the silhouettes and not the small surface details which will be washed out by the shading in the end. Most game studios prefer evenly spaced polygons for this and tend to cover large areas with very few edges - they want to focus the poly detail on the highly curved parts instead.
Second, anatomy shouldn’t be treated as static, one always have to keep movement in mind with characters. And there’s a lot of movement under the skin here, clavicles and scapulas and rib cages move around a lot. The way you place your edges based on that single pose can not accomodate the range of movement, the vast changes in shape that a real living creature goes through. This is mostly why every studio prefers to work with a lot of even geometry because they want to separate the anatomical structures and the flexible skin that covers them.
Localizing geometry to add detail seems to be your biggest concern as it does allow for more articulation to occur. I assure you this models topology can be turned to a higher poly mesh with more anatomical detail, with cuts and bevels. But the base form of the body deforms pretty well, since it was used in a training video. Except the forearm, its one area, I feel
Localizing is actually quite rare, we tend to remove the edges from the cranium to keep the model a bit lighter because that area is going to be covered with hair anyway. It does make the surface a bit bumpy so we don’t use it on other areas as much.
The models are actually quite similar to NURBS patch models of the previous decade, with the exception that we can place the poles at irregular intervals and we don’t have to pay computational and other costs to keep the patch edges stitched together, there are no cracks and we can UV map the model in whatever way we want, too.
It would be nice to push the boundaries in models in the 350K range but at some of the locations I taught it would crash the computers
There’s quite a range between 7,000 and 350,000 and I’m sure that there’s room for various compromises. I merely pointed out that staying that low isn’t a good idea nowadays. Also, your models are clearly subdivided which isn’t really possible in games.
The image you posted earlier, I do like the heavy use of localization in the face. I am curious to know, so I can stay up on industry standards, how did you build it out?
I’d like to note that I merely try to keep up myself, in no way do I consider our stuff to be at the front. We don’t have schedules and budgets to be the leaders in anything, we just try to learn and follow as best as we can.
Now the modeling method is simple - we start with concept sculpts for most organic stuff where an artist defines the forms and shapes, and then the modelers come in and build the geometry on top of that sculpt. Maya has Nex, 3ds max has Wrapit and the Graphite tools, there’s Topogun, 3D Coat, even Zbrush.
This way we can concentrate on the mesh itself as the artistic decisions have already been made, and the approach also makes it easy to completely rework any area without loosing the shape.
I see no reason not to start with Zbrush 3D concept sculpts any more, it just makes sense in every possible way. No wonder everyone’s doing it, even for hard surface with the new brushes in R4.