Implicit conversion from float -9.72270363951472 to int loses precision