Implicit conversion from float -101.41304347826085 to int loses precision