I wrote up a blog post about the strengths of each endianness:

https://technicalsourcery.net/posts/on-endianness/

Short short version:

It's all about where your imaginary "margin" is. Little endian works best with "upward growing" data, where the data grows from the low bits into the empty high bits (for example integers). Big endian works best with "downward growing" data, where the data grows from the high bits to the empty low bits (for example fractional portions).

Little endian has the biggest advantages because we use integer types a LOT more than floating point types.