Implicit conversion from float -98.54748603351956 to int loses precision