I think both are bad if you don't do any strength calculations
Generally speaking the predominant load in a low speed supercharged application will be bending so the last thing you want is a stress raiser where the crank pin meets the cheek. In this particular instance you probably want a hirth joint in the middle of the pin which implies you'll have to use some form of rolling element bearing.
A high speed N/A engine crank going through torsional resonance will be limited by the torque capacity of the joint regardless of where that is on pin.
Why not use a billet crankshaft? If you are planning on using journal bearings there are many machining schemes that would allow you to hollow out the pin and it will be stronger for the same general dimensions.
Finally, on polygon interference profiles: no matter how precisely you machine the internal and external forms you will always end up with a significant profile deviation on both sides compared to a simple circular interference. As a result, the interference pressure will always have large variations across the profile which means localized fretting is very likely.
Finally, to interfere a polygon you have to do it on a taper such that the actual interference is controlled by the press distance.