Implicit conversion from float -4.47115384615384 to int loses precision