the huntsman arrows use the collision hull, rather than the hitboxes

As far as I can tell it does. See this video.

The final arrow location isn’t directly representative of where the arrow hit, since it always goes through the center of a player bone.

Though I suppose that’s not the author’s fault. So nice collation.

Still, in relation to the actual article content, I don’t understand why my backstabs on my home computer are so different to the backstabs in town on a lan or those described here. For some reason, my backstabs act just like they did pre-instastab patch. When the knife falls, so does the victim. I can turn my knife mid-swing and everything. Yet in town, people just fall over dead as soon as I press the button.
Something weird is going on here with lag or latency, but I have no idea what.

My ideas focused on the fact that the playermodel stands with a twisted (to the LOS) back, and without any idea of the “collision hull” I assumed that to compensate the developers counted the back AND outwards-facing sides of the right-arm (which is somewhat behind the playermodel on most characters, as it is the “trigger” hand and does not extend as far as the other, which steadies the weapon) as the back for backstabs. This explained the visual phenomena of the sidestab, which APPEAR to occur from the right side because of the twisting of the playermodel.

Of course thats all moot now, I just thought it’s interesting how simply your model explains my own seperate observation and concern (the apparent ease of sidestabs from the right (spy left)), compared to my own- a sign that you have undoubtedly found the true mechanic!

Looking forward to more articles about the intricacies of TF2 (if there are going to be any, I for one would love it).

Anyway, that would explain why crouching is important when going for a stab from above.

Reid: I do want to mention that “true” facestabs–ones that are wrong even from a technical perspective–are real and have been occasionally reproduced. But the majority are probably “false” ones caused by lag-compensation which don’t look wrong to the attacking spy.

News: Not sure. Probably legacy reasons, and it might lead to trickier physics. An octagonal prism could have been a good compromise.

Lia: There’s some good information on the Valve Developer Community Wiki.

Curley: Not sure. I’d guess most are identical, but there’s no reason Valve couldn’t have programmed different weapons to have a different ranges if they wanted. It should be pretty straightforward to test with bots.

_____________

I’m playing with some old theories about “true” facestabs and looking over some of the diagrams I made, but in a way what I really need to do is make it an animation or slide-based flipbook because it gets confusing. You’ve got the client’s point-of-view, the server’s point of view, and the server’s guess at at what the client saw… and players moving through time and along a line.