This is pretty inconsistent. You should really specify to how many decimal places rounding needs to take place. If you're going to disregard 0.87+s and round it down, why is it that 0.999992 requires you to all of a sudden round up? You're basically creating edge cases that follow different rounding rules. It only affects certain solutions but that should still be an issue. Solutions that should work fail tests because of this gap in logic.