Implicit conversion from float -95.86956521739128 to int loses precision