It has been debated to death (and proved in all shapes and forms): "luck" is not affected if you scan either sequentially vs. some picked untested key (either at a random position, or via whatever other made-up logic, like prefixes).

What gets affected though, is the efficiency of the computations, and there is absolutely nothing currently that is known to be more efficient (e.g. that costs less or runs faster) than a full-on sequential scan. That is, if one wants to solve anything in the most efficient manner possible, not playing bingo with the elliptic curve and two uniform hash functions.