FWIW, I hadn't noticed this change in py3, but it would never have occurred to me that it's controversial... I thought everyone who'd studied the issue agreed that round-half-to-even was the best default :-). Numpy has always done round-to-even, and it's the default rounding mode mandated by IEEE754. (This doesn't mean that they mandate that round() itself follow this rule, but the implicit rounding that all floating point operations do has to follow this rule.) See also: https://en.wikipedia.org/wiki/Rounding#Round_half_to_even