When working with the excess representation of integers, I use a bias of 2n-1. However, the IEEE 754 standard instead uses 2n-1 - 1.
The only benefit that I can think of is a bigger positive range. Are there any other reasons as to why that decision was taken?