I misread the rules for Shift (not the document's fault), and assumed that the cards re-wrapped around (so when shifting right the rightmost card will be moved all the way to the left.  Still an interesting variant though, I think it combines the new complexity of shift with the compactness of the base game.  Preliminary testing: Averaging 12 million states per game (compared to closer to 1 million per for the non-shift variant.  And also, academically, a 90-10 bias for player 1.  Not confident on any of these numbers, as the additional complexity makes it much harder to verify computationally.  Hoping to do some more work on collapsi, just life getting in the way.  Cheers :)