Some things differs definitely, some not.

PS: just found out that US versions are PBN instead, but they;re working as well (they have different content though)

- I guess the vertices are the same and mostly for polys, though we'll only know once we progressed.
(from what I've seen with that little app that renders as bitmap, i.e. PCM sounds are instantly recognisable and those that I suspect being the above twos do look like WO1/2)

- there is an object system with names (cubeXX, goteki, etc ...) my initial thoughts were to attack the format by studying the 2nd prototype track as it's very simple and contains cubes, these are small structs that once we figure one of them we should progress quite a bit. I think I've figured where are the indices but it's a bit puzzling.

- i do not suspect other things to differ that much since as we're on a PS1 and as you can see in no$psx, the rendering possibilities are known, there aren't 100s of possible combinations (tri, quad, face, vert, etc ...)

This in conjunction with a vertices renderer should significantly help, basically if you render an array of vertices as spheres and look at them from a certain point of view you'd know very quickly if's vertex data or not; e.g. the 2nd prototype track will be of help because there are orphan cubes in the circuit, a significant hint. (will test that theory with WOXL data first as we know what to expect)

for the rest you'll discover that by yourself

|
|
V

UnpackPBP is done and works well, still would like to throw all the pb(n/p)s to test it, writing a little doc, a few other things then I'll put on GitHub.

It's been quite time consuming to get from a pseudo-format to a well understood (i guess by now) format. Maybe tonight or tomorrow, actually time runs pretty fast when you do something interesting !