[danl1]

Shimano explicitly recommends a particular fore-aft direction in their instructions - and strongly recommends it. (See note on right of page)

http://techdocs.shimano.com/media/techdocs/content/cycle/SI/SI_08V0C/SI_08V0C_13_v1_m56577569830612411.pdf

And while Campy doesn't mention it explicitly, following their instructions has the same effect:

http://www.campagnolo.com/repository/documenti/en/7225362-11s_chains.pdf

Also, they each have ideas about which way the joining pin should be pushed through. That represents a 'which side' direction in practice, at least to the finished loop of chain.

As we move to 11 speeds, there's talk of differential shaping of the outer plates, as space is becoming a premium and the demands of shifting to larger cogs are different from shfting to larger rings. We'll see where that goes.

 

[danl1]

Thanks danl1. Can someone explain the engineering behind this recommendation? Looking at the diagram, Shimano seems to recommend the inner link pulling the outer link at the pin. Why would this be advantageous to the other way around? It seems counter intuitive, as I would think it would have a tendency to split apart the outer link from the pin. Actually, come to think of it, why would it make a difference? The chain doesn't know which side it's being pulled on.

Honestly don't know - just spitballing here. You are right that the force is symmetrical - the chain doesn't know if the rings or cogs are doing the pulling.

My guess is that it comes into play when cross-chaining to some degree. Since the narrower gap in the inner plates disengages/engages the teeth first, the extra room between the outer plates allows the engaged/engaging tooth to move a bit, allowing the lateral bend to happen gradually, or at least elsewhere than at this weakest pin joint. Routed the other way, the narrow plates 'grab' the tooth, creating a force more concentrated on that joint as it dis/engages. That would still be a symmetrical problem, except that the chain wraps tighter around the cogs than the chainrings, and so the disengagement torques the chain more than the engagement.

Another possibility might have to do with the way the chain hits ramps and pins when shifting, but it'd take more effort to BS my way thru that one than I'm willing to commit right now.