It's also called banker's rounding. To understand why it's a good default, you can check out https://mathematica.stackexchange.com/a/2120

The motivation for the seemingly strange rounding is that a set of numbers can be rounded before summing without biasing the sum. It has good use cases but so do all rounding modes.

In general, if some recommendation from IEEE 754 (the floating-point standard) seems strange, its because someone like William Kahan thought about it in great detail and you'd be wise to follow it.

https://en.wikipedia.org/wiki/Rounding#Rounding_half_to_even