How to achieve similar, clean topology?

Hi, totally newbie(less than 6months in 3D in general) so be gentle please

Yesterday in web i found this chair(on pics below), and was amazed about its topology. Whole chair is about 26.500 faces which is quite low, i think.

But i cant understand how to make it? I know i can use Sculpt mode to make folds and wrinkles. Which i did for my chair. Which took about 1.300.000 faces in hi-poly. And 20.000 low poly. So i used hi-poly to make normal map and bake textures(for game).

Question is - how to achieve pretty same quality and topology for low poly cost? And no triangulation as dyntopo.
Thank you!

The question is quite weird. Youâ€™ve done what you then ask how to do. Itâ€™s also weird because a low poly game model can use triangles to optimize polygon count. Itâ€™s also a good idea to triangulate before baking the normal map so that the geometry doesnâ€™t change between applications. If the other application would triangulate it differently, the baked normals and underlying geometry wouldnâ€™t match.

If youâ€™re asking how to do the low poly version more efficiently, saving more polygons, thereâ€™s no magic to it. Pay attention to silhouette because that has to be defined with real geometry. Otherwise a normal map can make it look quite detailed with less polygons. Hereâ€™s a good article on that http://cgi.tutsplus.com/articles/game-character-creation-series-kila-chapter-2-low-resolution-modeling--cg-25788

My bad, probably misunderstood question. Ill try to ask more precisely. Main question is - how to achieve topology as on 1st picture in topic. If you pay attention, topology flowing/following wrinkles. + Nice control of polygons density, more dense on wrinkles to make them smoother and less in flatter parts. All this together helps to make model look like its almost high poly(sculpted) but with relatively low poly count. Im trying to understand workflow, how to achieve that quality and control of density.

About triangles - im not game modeler, just learning modeling for myself so far and trying to model stuff not just for create it and forget, its not that interesting for me when make model and no where to apply. Im always try to make for reason, thats help to stay involved and understand what i wanna to get in the end. Andâ€¦ in the end when i export model i always triangulate it, sure But in blender as i know it always better to work with quads, dyntopo not good for that reason.

I hope now i ask better

Decide the level of detail you want to describe in real geometry, follow high points and crevises with edge loops, hard ones preferably with 3, and then start modeling. Planning ahead is important.

Subdivision might have been used in the example by first doing a bit rough tracing of the curves, apply subdivision, and then reducing it.